Daily Papers

As LLM agents tackle increasingly complex tasks, a critical question is whether they can maintain strategic coherence over long horizons: planning under uncertainty, learning from delayed feedback, and adapting when early mistakes compound. We introduce YC-Bench, a benchmark that evaluates these capabilities by tasking an agent with running a simulated startup over a one-year horizon spanning hundreds of turns. The agent must manage employees, select task contracts, and maintain profitability in a partially observable environment where adversarial clients and growing payroll create compounding consequences for poor decisions. We evaluate 12 models, both proprietary and open source, across 3 seeds each. Only three models consistently surpass the starting capital of \200K, with Claude Opus 4.6 achieving the highest average final funds at 1.27 M, followed by GLM-5 at \1.21 M at 11\times lower inference cost. Scratchpad usage, the sole mechanism for persisting information across context truncation, is the strongest predictor of success, and adversarial client detection is the primary failure mode, accounting for 47\% of bankruptcies. Our analysis reveals that frontier models still fail through distinct failure modes such as over-parallelization, demonstrating the capability gaps for long-horizon performance. YC-Bench$ is open-source, reproducible, and configurable.

We introduce the Scratchpad Mechanism, a novel addition to the sequence-to-sequence (seq2seq) neural network architecture and demonstrate its effectiveness in improving the overall fluency of seq2seq models for natural language generation tasks. By enabling the decoder at each time step to write to all of the encoder output layers, Scratchpad can employ the encoder as a "scratchpad" memory to keep track of what has been generated so far and thereby guide future generation. We evaluate Scratchpad in the context of three well-studied natural language generation tasks --- Machine Translation, Question Generation, and Text Summarization --- and obtain state-of-the-art or comparable performance on standard datasets for each task. Qualitative assessments in the form of human judgements (question generation), attention visualization (MT), and sample output (summarization) provide further evidence of the ability of Scratchpad to generate fluent and expressive output.

Diffusion-based Large Language Models (dLLMs) parallelize text generation by framing decoding as a denoising process, but suffer from high computational overhead since they predict all future suffix tokens at each step while retaining only a small fraction. We propose Diffusion Scratchpad (DPad), a training-free method that restricts attention to a small set of nearby suffix tokens, preserving fidelity while eliminating redundancy. DPad integrates two strategies: (i) a sliding window, which maintains a fixed-length suffix window, and (ii) distance-decay dropout, which deterministically removes distant suffix tokens before attention computation. This simple design is compatible with existing optimizations such as prefix caching and can be implemented with only a few lines of code. Comprehensive evaluations across multiple benchmarks on LLaDA-1.5 and Dream models demonstrate that DPad delivers up to 61.4times speedup over vanilla dLLMs while maintaining comparable accuracy, highlighting its potential for efficient and scalable long-sequence inference. Our code is available at https://github.com/Crys-Chen/DPad.

Large pre-trained language models perform remarkably well on tasks that can be done "in one pass", such as generating realistic text or synthesizing computer programs. However, they struggle with tasks that require unbounded multi-step computation, such as adding integers or executing programs. Surprisingly, we find that these same models are able to perform complex multi-step computations -- even in the few-shot regime -- when asked to perform the operation "step by step", showing the results of intermediate computations. In particular, we train transformers to perform multi-step computations by asking them to emit intermediate computation steps into a "scratchpad". On a series of increasingly complex tasks ranging from long addition to the execution of arbitrary programs, we show that scratchpads dramatically improve the ability of language models to perform multi-step computations.

Humans draw to facilitate reasoning: we draw auxiliary lines when solving geometry problems; we mark and circle when reasoning on maps; we use sketches to amplify our ideas and relieve our limited-capacity working memory. However, such actions are missing in current multimodal language models (LMs). Current chain-of-thought and tool-use paradigms only use text as intermediate reasoning steps. In this work, we introduce Sketchpad, a framework that gives multimodal LMs a visual sketchpad and tools to draw on the sketchpad. The LM conducts planning and reasoning according to the visual artifacts it has drawn. Different from prior work, which uses text-to-image models to enable LMs to draw, Sketchpad enables LMs to draw with lines, boxes, marks, etc., which is closer to human sketching and better facilitates reasoning. Sketchpad can also use specialist vision models during the sketching process (e.g., draw bounding boxes with object detection models, draw masks with segmentation models), to further enhance visual perception and reasoning. We experiment with a wide range of math tasks (including geometry, functions, graphs, and chess) and complex visual reasoning tasks. Sketchpad substantially improves performance on all tasks over strong base models with no sketching, yielding an average gain of 12.7% on math tasks, and 8.6% on vision tasks. GPT-4o with Sketchpad sets a new state of the art on all tasks, including V*Bench (80.3%), BLINK spatial reasoning (83.9%), and visual correspondence (80.8%). All codes and data are in https://visualsketchpad.github.io/.

Tokenizer-free language models eliminate the tokenizer step of the language modeling pipeline by operating directly on bytes; patch-based variants further aggregate contiguous byte spans into patches for efficiency. However, the average patch size chosen at the model design stage governs a tight trade-off: larger patches reduce compute and KV-cache footprint, but degrade modeling quality. We trace this trade-off to patch lag: until a patch is fully observed, byte predictions within it must rely on a stale representation from the previous patch to preserve causality; this lag widens as patches grow larger. We introduce Scratchpad Patching (SP), which inserts transient scratchpads inside each patch to aggregate the bytes seen so far and refresh patch-level context for subsequent predictions. SP triggers scratchpads using next-byte prediction entropy, selectively allocating compute to information-dense regions and enabling post-hoc adjustment of inference-time compute. Across experiments on natural language and code, SP improves model quality at the same patch size; for example, even at 16 bytes per patch, SP-augmented models match or closely approach the byte-level baseline on downstream evaluations while using a 16times smaller KV cache over patches and 3-4times less inference compute.

Large language models have been shown to struggle with limited context memory and multi-step reasoning. We propose a simple method for solving both of these problems by allowing the model to take Self-Notes. Unlike recent scratchpad approaches, the model can deviate from the input context at any time to explicitly think. This allows the model to recall information and perform reasoning on the fly as it reads the context, thus extending its memory and enabling multi-step reasoning. Our experiments on multiple tasks demonstrate that our method can successfully generalize to longer and more complicated instances from their training setup by taking Self-Notes at inference time.

Recent advancements in large multimodal models (LMMs) have showcased impressive code generation capabilities, primarily evaluated through image-to-code benchmarks. However, these benchmarks are limited to specific visual programming scenarios where the logic reasoning and the multimodal understanding capacities are split apart. To fill this gap, we propose ScratchEval, a novel benchmark designed to evaluate the visual programming reasoning ability of LMMs. ScratchEval is based on Scratch, a block-based visual programming language widely used in children's programming education. By integrating visual elements and embedded programming logic, ScratchEval requires the model to process both visual information and code structure, thereby comprehensively evaluating its programming intent understanding ability. Our evaluation approach goes beyond the traditional image-to-code mapping and focuses on unified logical thinking and problem-solving abilities, providing a more comprehensive and challenging framework for evaluating the visual programming ability of LMMs. ScratchEval not only fills the gap in existing evaluation methods, but also provides new insights for the future development of LMMs in the field of visual programming. Our benchmark can be accessed at https://github.com/HKBUNLP/ScratchEval .

Can Transformers predict new syllogisms by composing established ones? More generally, what type of targets can be learned by such models from scratch? Recent works show that Transformers can be Turing-complete in terms of expressivity, but this does not address the learnability objective. This paper puts forward the notion of 'globality degree' of a target distribution to capture when weak learning is efficiently achievable by regular Transformers, where the latter measures the least number of tokens required in addition to the tokens histogram to correlate nontrivially with the target. As shown experimentally and theoretically under additional assumptions, distributions with high globality cannot be learned efficiently. In particular, syllogisms cannot be composed on long chains. Furthermore, we show that (i) an agnostic scratchpad cannot help to break the globality barrier, (ii) an educated scratchpad can help if it breaks the globality at each step, however not all such scratchpads can generalize to out-of-distribution (OOD) samples, (iii) a notion of 'inductive scratchpad', that composes the prior information more efficiently, can both break the globality barrier and improve the OOD generalization. In particular, some inductive scratchpads can achieve length generalizations of up to 6x for some arithmetic tasks depending on the input formatting.

The ability to extrapolate from short problem instances to longer ones is an important form of out-of-distribution generalization in reasoning tasks, and is crucial when learning from datasets where longer problem instances are rare. These include theorem proving, solving quantitative mathematics problems, and reading/summarizing novels. In this paper, we run careful empirical studies exploring the length generalization capabilities of transformer-based language models. We first establish that naively finetuning transformers on length generalization tasks shows significant generalization deficiencies independent of model scale. We then show that combining pretrained large language models' in-context learning abilities with scratchpad prompting (asking the model to output solution steps before producing an answer) results in a dramatic improvement in length generalization. We run careful failure analyses on each of the learning modalities and identify common sources of mistakes that highlight opportunities in equipping language models with the ability to generalize to longer problems.

Assessing student handwritten scratchwork is crucial for personalized educational feedback but presents unique challenges due to diverse handwriting, complex layouts, and varied problem-solving approaches. Existing educational NLP primarily focuses on textual responses and neglects the complexity and multimodality inherent in authentic handwritten scratchwork. Current multimodal large language models (MLLMs) excel at visual reasoning but typically adopt an "examinee perspective", prioritizing generating correct answers rather than diagnosing student errors. To bridge these gaps, we introduce ScratchMath, a novel benchmark specifically designed for explaining and classifying errors in authentic handwritten mathematics scratchwork. Our dataset comprises 1,720 mathematics samples from Chinese primary and middle school students, supporting two key tasks: Error Cause Explanation (ECE) and Error Cause Classification (ECC), with seven defined error types. The dataset is meticulously annotated through rigorous human-machine collaborative approaches involving multiple stages of expert labeling, review, and verification. We systematically evaluate 16 leading MLLMs on ScratchMath, revealing significant performance gaps relative to human experts, especially in visual recognition and logical reasoning. Proprietary models notably outperform open-source models, with large reasoning models showing strong potential for error explanation. All evaluation data and frameworks are publicly available to facilitate further research.

While Multimodal Large Language Models (MLLMs) excel at visual understanding, they often struggle in complex scenarios that require visual planning and imagination. Inspired by how humans use sketching as a form of visual thinking to develop and communicate ideas, we introduce Latent Sketchpad, a framework that equips MLLMs with an internal visual scratchpad. The internal visual representations of MLLMs have traditionally been confined to perceptual understanding. We repurpose them to support generative visual thought without compromising reasoning ability. Building on frontier MLLMs, our approach integrates visual generation directly into their native autoregressive reasoning process. It allows the model to interleave textual reasoning with the generation of visual latents. These latents guide the internal thought process and can be translated into sketch images for interpretability. To realize this, we introduce two components: a Context-Aware Vision Head autoregressively produces visual representations, and a pretrained Sketch Decoder renders these into human-interpretable images. We evaluate the framework on our new dataset MazePlanning. Experiments across various MLLMs show that Latent Sketchpad delivers comparable or even superior reasoning performance to their backbone. It further generalizes across distinct frontier MLLMs, including Gemma3 and Qwen2.5-VL. By extending model's textual reasoning to visual thinking, our framework opens new opportunities for richer human-computer interaction and broader applications. More details and resources are available on our project page: https://latent-sketchpad.github.io/.

Graphical user interface (GUI) grounding, the process of mapping human instructions to GUI actions, serves as a fundamental basis to autonomous GUI agents. While existing grounding models achieve promising performance to simulate the mouse click action on various click-based benchmarks, another essential mode of mouse interaction, namely dragging, remains largely underexplored. Yet, dragging the mouse to select and manipulate textual content represents a prevalent and important usage in practical GUI scenarios. To narrow this gap, we first introduce GUI-Drag, a diverse dataset of 161K text dragging examples synthesized through a scalable pipeline. To support systematic and robust evaluation, we further construct ScreenDrag, a benchmark with 5,333 examples spanning three levels of interface context, together with three dedicated metrics designed for assessing text dragging capability. Models trained on GUI-Drag with an efficient continual training strategy achieve substantial improvements on ScreenDrag, while preserving the original click-based performance on ScreenSpot, ScreenSpot-v2, and OSWorld-G. Our work encourages further research on broader GUI grounding beyond just clicking and paves way toward a truly generalist GUI grounding model. All benchmark, data, checkpoints, and code are open-sourced and available at https://osu-nlp-group.github.io/GUI-Drag.

Inference-time scaling can enhance the reasoning capabilities of large language models (LLMs) on complex problems that benefit from step-by-step problem solving. Although lengthening generated scratchpads has proven effective for mathematical tasks, the broader impact of this approach on other tasks remains less clear. In this work, we investigate the benefits and limitations of scaling methods across nine state-of-the-art models and eight challenging tasks, including math and STEM reasoning, calendar planning, NP-hard problems, navigation, and spatial reasoning. We compare conventional models (e.g., GPT-4o) with models fine-tuned for inference-time scaling (e.g., o1) through evaluation protocols that involve repeated model calls, either independently or sequentially with feedback. These evaluations approximate lower and upper performance bounds and potential for future performance improvements for each model, whether through enhanced training or multi-model inference systems. Our extensive empirical analysis reveals that the advantages of inference-time scaling vary across tasks and diminish as problem complexity increases. In addition, simply using more tokens does not necessarily translate to higher accuracy in these challenging regimes. Results from multiple independent runs with conventional models using perfect verifiers show that, for some tasks, these models can achieve performance close to the average performance of today's most advanced reasoning models. However, for other tasks, a significant performance gap remains, even in very high scaling regimes. Encouragingly, all models demonstrate significant gains when inference is further scaled with perfect verifiers or strong feedback, suggesting ample potential for future improvements.

Interacting with Large Language Models (LLMs) for text editing on mobile devices currently requires users to break out of their writing environment and switch to a conversational AI interface. In this paper, we propose to control the LLM via touch gestures performed directly on the text. We first chart a design space that covers fundamental touch input and text transformations. In this space, we then concretely explore two control mappings: spread-to-generate and pinch-to-shorten, with visual feedback loops. We evaluate this concept in a user study (N=14) that compares three feedback designs: no visualisation, text length indicator, and length + word indicator. The results demonstrate that touch-based control of LLMs is both feasible and user-friendly, with the length + word indicator proving most effective for managing text generation. This work lays the foundation for further research into gesture-based interaction with LLMs on touch devices.

Modern vision models have achieved remarkable success in benchmarks where local features provide critical information about the target. There is now a growing interest in solving tasks that require more global reasoning, where local features offer no significant information. These tasks are reminiscent of the connectivity tasks discussed by Minsky and Papert in 1969, which exposed the limitations of the perceptron model and contributed to the first AI winter. In this paper, we revisit such tasks by introducing four global visual benchmarks involving path findings and mazes. We show that: (1) although today's large vision models largely surpass the expressivity limitations of the early models, they still struggle with the learning efficiency; we put forward the "globality degree" notion to understand this limitation; (2) we then demonstrate that the picture changes and global reasoning becomes feasible with the introduction of "visual scratchpads"; similarly to the text scratchpads and chain-of-thoughts used in language models, visual scratchpads help break down global tasks into simpler ones; (3) we finally show that some scratchpads are better than others, in particular, "inductive scratchpads" that take steps relying on less information afford better out-of-distribution generalization and succeed for smaller model sizes.

Large Reasoning Models (LRMs) have become a central focus in today's large language model (LLM) research, where models are designed to output a step-by-step thinking process before arriving at a final answer to handle complex reasoning tasks. Despite their promise, recent empirical studies (e.g., [Shojaee et al., 2025] from Apple) suggest that this thinking process may not actually enhance reasoning ability, where LLMs without explicit reasoning actually outperform LRMs on tasks with low or high complexity. In this work, we revisit these findings and investigate whether the limitations of LRMs persist when tool augmentations are introduced. We incorporate two types of tools, Python interpreters and scratchpads, and evaluate three representative LLMs and their LRM counterparts on Apple's benchmark reasoning puzzles. Our results show that, with proper tool use, LRMs consistently outperform their non-reasoning counterparts across all levels of task complexity. These findings challenge the recent narrative that reasoning is an illusion and highlight the potential of tool-augmented LRMs for solving complex problems.

Large language models (LLMs) have become an essential tool to support developers using traditional text-based programming languages, but the graphical notation of the block-based Scratch programming environment inhibits the use of LLMs. To overcome this limitation, we propose the LitterBox+ framework that extends the Scratch static code analysis tool LitterBox with the generative abilities of LLMs. By converting block-based code to a textual representation suitable for LLMs, LitterBox+ allows users to query LLMs about their programs, about quality issues reported by LitterBox, and it allows generating code fixes. Besides offering a programmatic API for these functionalities, LitterBox+ also extends the Scratch user interface to make these functionalities available directly in the environment familiar to learners. The framework is designed to be easily extensible with other prompts, LLM providers, and new features combining the program analysis capabilities of LitterBox with the generative features of LLMs. We provide a screencast demonstrating the tool at https://youtu.be/RZ6E0xgrIgQ.

We introduce DrawTalking, an approach to building and controlling interactive worlds by sketching and speaking while telling stories. It emphasizes user control and flexibility, and gives programming-like capability without requiring code. An early open-ended study with our prototype shows that the mechanics resonate and are applicable to many creative-exploratory use cases, with the potential to inspire and inform research in future natural interfaces for creative exploration and authoring.

Digital note-taking is gaining popularity, offering a durable, editable, and easily indexable way of storing notes in the vectorized form, known as digital ink. However, a substantial gap remains between this way of note-taking and traditional pen-and-paper note-taking, a practice still favored by a vast majority. Our work, InkSight, aims to bridge the gap by empowering physical note-takers to effortlessly convert their work (offline handwriting) to digital ink (online handwriting), a process we refer to as Derendering. Prior research on the topic has focused on the geometric properties of images, resulting in limited generalization beyond their training domains. Our approach combines reading and writing priors, allowing training a model in the absence of large amounts of paired samples, which are difficult to obtain. To our knowledge, this is the first work that effectively derenders handwritten text in arbitrary photos with diverse visual characteristics and backgrounds. Furthermore, it generalizes beyond its training domain into simple sketches. Our human evaluation reveals that 87% of the samples produced by our model on the challenging HierText dataset are considered as a valid tracing of the input image and 67% look like a pen trajectory traced by a human.

Sketching serves as a versatile tool for externalizing ideas, enabling rapid exploration and visual communication that spans various disciplines. While artificial systems have driven substantial advances in content creation and human-computer interaction, capturing the dynamic and abstract nature of human sketching remains challenging. In this work, we introduce SketchAgent, a language-driven, sequential sketch generation method that enables users to create, modify, and refine sketches through dynamic, conversational interactions. Our approach requires no training or fine-tuning. Instead, we leverage the sequential nature and rich prior knowledge of off-the-shelf multimodal large language models (LLMs). We present an intuitive sketching language, introduced to the model through in-context examples, enabling it to "draw" using string-based actions. These are processed into vector graphics and then rendered to create a sketch on a pixel canvas, which can be accessed again for further tasks. By drawing stroke by stroke, our agent captures the evolving, dynamic qualities intrinsic to sketching. We demonstrate that SketchAgent can generate sketches from diverse prompts, engage in dialogue-driven drawing, and collaborate meaningfully with human users.

Teaching Computer Science (CS) by having students write programs by hand on paper has key pedagogical advantages: It allows focused learning and requires careful thinking compared to the use of Integrated Development Environments (IDEs) with intelligent support tools or "just trying things out". The familiar environment of pens and paper also lessens the cognitive load of students with no prior experience with computers, for whom the mere basic usage of computers can be intimidating. Finally, this teaching approach opens learning opportunities to students with limited access to computers. However, a key obstacle is the current lack of teaching methods and support software for working with and running handwritten programs. Optical character recognition (OCR) of handwritten code is challenging: Minor OCR errors, perhaps due to varied handwriting styles, easily make code not run, and recognizing indentation is crucial for languages like Python but is difficult to do due to inconsistent horizontal spacing in handwriting. Our approach integrates two innovative methods. The first combines OCR with an indentation recognition module and a language model designed for post-OCR error correction without introducing hallucinations. This method, to our knowledge, surpasses all existing systems in handwritten code recognition. It reduces error from 30\% in the state of the art to 5\% with minimal hallucination of logical fixes to student programs. The second method leverages a multimodal language model to recognize handwritten programs in an end-to-end fashion. We hope this contribution can stimulate further pedagogical research and contribute to the goal of making CS education universally accessible. We release a dataset of handwritten programs and code to support future research at https://github.com/mdoumbouya/codeocr

Block-based programming languages like Scratch are increasingly popular for programming education and end-user programming. Recent program analyses build on the insight that source code can be modelled using techniques from natural language processing. Many of the regularities of source code that support this approach are due to the syntactic overhead imposed by textual programming languages. This syntactic overhead, however, is precisely what block-based languages remove in order to simplify programming. Consequently, it is unclear how well this modelling approach performs on block-based programming languages. In this paper, we investigate the applicability of language models for the popular block-based programming language Scratch. We model Scratch programs using n-gram models, the most essential type of language model, and transformers, a popular deep learning model. Evaluation on the example tasks of code completion and bug finding confirm that blocks inhibit predictability, but the use of language models is nevertheless feasible. Our findings serve as foundation for improving tooling and analyses for block-based languages.

Graphical User Interface (GUI) tasks are vital for automating workflows such as software testing, user interface navigation. For users, the GUI is the most intuitive platform for interacting with a computer. Previous work identified a key challenge in developing visual GUI agents: GUI grounding - the ability to accurately locate screen elements based on instructions. However, most existing GUI agents rely on structured data formats like DOM or HTML files in training or inferencing, which are inaccessible across all applications, particular in a general desktop environments such as Windows OS. To address this, we introduce WinClick, a novel visual GUI agent developed in Windows platform. WinClick leverages screenshots to detect actionable regions. To overcome the challenge of GUI grounding, we enhance WinClick with GUI grounding pre-training and propose an LLM-based method for aligning GUI grounding data. Additionally, we introduce WinSpot, the first comprehensive benchmark for GUI grounding on Windows. Our experiments demonstrate that WinClick, combined with GUI grounding pre-training, significantly outperforms existing baselines, offering a scalable solution for GUI automation in desktop environments. WinSpot is publicly available at https://github.com/zackhuiiiii/WinSpot.

The task of inserting text into a specified position in a passage, known as fill in the blank (FitB), is useful for a variety of applications where writers interact with a natural language generation (NLG) system to craft text. While previous work has tackled this problem with models trained specifically to do the fill-in-the-blank task, a more useful model is one that can effectively perform _both_ FitB and continuation. In this work, we evaluate the feasibility of using a single model to do both tasks. We show that models pre-trained with a FitB-style objective are capable of both tasks, while models pre-trained for continuation are not. Finally, we show how FitB models can be easily finetuned to allow for fine-grained control over the length and word choice of the generation.

Length generalization, the ability to generalize from small training context sizes to larger ones, is a critical challenge in the development of Transformer-based language models. Positional encoding (PE) has been identified as a major factor influencing length generalization, but the exact impact of different PE schemes on extrapolation in downstream tasks remains unclear. In this paper, we conduct a systematic empirical study comparing the length generalization performance of decoder-only Transformers with five different position encoding approaches including Absolute Position Embedding (APE), T5's Relative PE, ALiBi, and Rotary, in addition to Transformers without positional encoding (NoPE). Our evaluation encompasses a battery of reasoning and mathematical tasks. Our findings reveal that the most commonly used positional encoding methods, such as ALiBi, Rotary, and APE, are not well suited for length generalization in downstream tasks. More importantly, NoPE outperforms other explicit positional encoding methods while requiring no additional computation. We theoretically demonstrate that NoPE can represent both absolute and relative PEs, but when trained with SGD, it mostly resembles T5's relative PE attention patterns. Finally, we find that scratchpad is not always helpful to solve length generalization and its format highly impacts the model's performance. Overall, our work suggests that explicit position embeddings are not essential for decoder-only Transformers to generalize well to longer sequences.

The needle-in-a-haystack (NIAH) test, which examines the ability to retrieve a piece of information (the "needle") from long distractor texts (the "haystack"), has been widely adopted to evaluate long-context language models (LMs). However, this simple retrieval-based test is indicative of only a superficial form of long-context understanding. To provide a more comprehensive evaluation of long-context LMs, we create a new synthetic benchmark RULER with flexible configurations for customized sequence length and task complexity. RULER expands upon the vanilla NIAH test to encompass variations with diverse types and quantities of needles. Moreover, RULER introduces new task categories multi-hop tracing and aggregation to test behaviors beyond searching from context. We evaluate ten long-context LMs with 13 representative tasks in RULER. Despite achieving nearly perfect accuracy in the vanilla NIAH test, all models exhibit large performance drops as the context length increases. While these models all claim context sizes of 32K tokens or greater, only four models (GPT-4, Command-R, Yi-34B, and Mixtral) can maintain satisfactory performance at the length of 32K. Our analysis of Yi-34B, which supports context length of 200K, reveals large room for improvement as we increase input length and task complexity. We open source RULER to spur comprehensive evaluation of long-context LMs.

Text-to-image generation models have grown in popularity due to their ability to produce high-quality images from a text prompt. One use for this technology is to enable the creation of more accessible art creation software. In this paper, we document the development of an alternative user interface that reduces the typing effort needed to enter image prompts by providing suggestions from a large language model, developed through iterative design and testing within the project team. The results of this testing demonstrate how generative text models can support the accessibility of text-to-image models, enabling users with a range of abilities to create visual art.

Documents serve as a crucial and indispensable medium for everyday workplace tasks. However, understanding, interacting and creating such documents on today's planar interfaces without any intelligent support are challenging due to our natural cognitive constraints on remembering, processing, understanding and interacting with these information. My doctorate research investigates how to bring 2D document interactions into immersive information experience using multiple of today's emergent technologies. With the examples of four specific types of documents -- medical scans, instruction document, self-report diary survey, and reference images for visual artists -- my research demonstrates how to transform such of today's 2D document interactions into an immersive information experience, by augmenting content with virtual reality, spatializing document placements with mixed reality, enriching long-term and continuous interactions with voice assistants, and simplify document creation workflow with generative AI.

Sketching provides an intuitive way to convey dynamic intent in animation authoring (i.e., how elements change over time and space), making it a natural medium for automatic content creation. Yet existing approaches often constrain sketches to fixed command tokens or predefined visual forms, overlooking their freeform nature and the central role of humans in shaping intention. To address this, we introduce an interaction paradigm where users convey dynamic intent to a vision-language model via free-form sketching, instantiated here in a sketch storyboard to motion graphics workflow. We implement an interface and improve it through a three-stage study with 24 participants. The study shows how sketches convey motion with minimal input, how their inherent ambiguity requires users to be involved for clarification, and how sketches can visually guide video refinement. Our findings reveal the potential of sketch and AI interaction to bridge the gap between intention and outcome, and demonstrate its applicability to 3D animation and video generation.

This paper presents Dynamic Context Prompting/Programming (DCP/P), a novel framework for interacting with LLMs to generate graph-based content with a dynamic context window history. While there is an existing study utilizing LLMs to generate a visual novel game, the previous study involved a manual process of output extraction and did not provide flexibility in generating a longer, coherent story. We evaluate DCP/P against our baseline, which does not provide context history to an LLM and only relies on the initial story data. Through objective evaluation, we show that simply providing the LLM with a summary leads to a subpar story compared to additionally providing the LLM with the proper context of the story. We also provide an extensive qualitative analysis and discussion. We qualitatively examine the quality of the objectively best-performing generated game from each approach. In addition, we examine biases in word choices and word sentiment of the generated content. We find a consistent observation with previous studies that LLMs are biased towards certain words, even with a different LLM family. Finally, we provide a comprehensive discussion on opportunities for future studies.

The proliferation of Large Language Models (LLMs) highlights the critical importance of conducting thorough evaluations to discern their comparative advantages, limitations, and optimal use cases. Particularly important is assessing their capacity to accurately retrieve information included in a given prompt. A model's ability to do this significantly influences how effectively it can utilize contextual details, thus impacting its practical efficacy and dependability in real-world applications. Our research analyzes the in-context recall performance of various LLMs using the needle-in-a-haystack method. In this approach, a factoid (the "needle") is embedded within a block of filler text (the "haystack"), which the model is asked to retrieve. We assess the recall performance of each model across various haystack lengths and with varying needle placements to identify performance patterns. This study demonstrates that an LLM's recall capability is not only contingent upon the prompt's content but also may be compromised by biases in its training data. Conversely, adjustments to model architecture, training strategy, or fine-tuning can improve performance. Our analysis provides insight into LLM behavior, offering direction for the development of more effective applications of LLMs.

Sketch animations offer a powerful medium for visual storytelling, from simple flip-book doodles to professional studio productions. While traditional animation requires teams of skilled artists to draw key frames and in-between frames, existing automation attempts still demand significant artistic effort through precise motion paths or keyframe specification. We present FlipSketch, a system that brings back the magic of flip-book animation -- just draw your idea and describe how you want it to move! Our approach harnesses motion priors from text-to-video diffusion models, adapting them to generate sketch animations through three key innovations: (i) fine-tuning for sketch-style frame generation, (ii) a reference frame mechanism that preserves visual integrity of input sketch through noise refinement, and (iii) a dual-attention composition that enables fluid motion without losing visual consistency. Unlike constrained vector animations, our raster frames support dynamic sketch transformations, capturing the expressive freedom of traditional animation. The result is an intuitive system that makes sketch animation as simple as doodling and describing, while maintaining the artistic essence of hand-drawn animation.

We introduce MeshPad, a generative approach that creates 3D meshes from sketch inputs. Building on recent advances in artistic-designed triangle mesh generation, our approach addresses the need for interactive artistic mesh creation. To this end, we focus on enabling consistent edits by decomposing editing into 'deletion' of regions of a mesh, followed by 'addition' of new mesh geometry. Both operations are invoked by simple user edits of a sketch image, facilitating an iterative content creation process and enabling the construction of complex 3D meshes. Our approach is based on a triangle sequence-based mesh representation, exploiting a large Transformer model for mesh triangle addition and deletion. In order to perform edits interactively, we introduce a vertex-aligned speculative prediction strategy on top of our additive mesh generator. This speculator predicts multiple output tokens corresponding to a vertex, thus significantly reducing the computational cost of inference and accelerating the editing process, making it possible to execute each editing step in only a few seconds. Comprehensive experiments demonstrate that MeshPad outperforms state-of-the-art sketch-conditioned mesh generation methods, achieving more than 22% mesh quality improvement in Chamfer distance, and being preferred by 90% of participants in perceptual evaluations.

Learning how humans manipulate objects requires machines to acquire knowledge from two perspectives: one for understanding object affordances and the other for learning human's interactions based on the affordances. Even though these two knowledge bases are crucial, we find that current databases lack a comprehensive awareness of them. In this work, we propose a multi-modal and rich-annotated knowledge repository, OakInk, for visual and cognitive understanding of hand-object interactions. We start to collect 1,800 common household objects and annotate their affordances to construct the first knowledge base: Oak. Given the affordance, we record rich human interactions with 100 selected objects in Oak. Finally, we transfer the interactions on the 100 recorded objects to their virtual counterparts through a novel method: Tink. The recorded and transferred hand-object interactions constitute the second knowledge base: Ink. As a result, OakInk contains 50,000 distinct affordance-aware and intent-oriented hand-object interactions. We benchmark OakInk on pose estimation and grasp generation tasks. Moreover, we propose two practical applications of OakInk: intent-based interaction generation and handover generation. Our datasets and source code are publicly available at https://github.com/lixiny/OakInk.

How can we better organize code in computational notebooks? Notebooks have become a popular tool among data scientists, as they seamlessly weave text and code together, supporting users to rapidly iterate and document code experiments. However, it is often challenging to organize code in notebooks, partially because there is a mismatch between the linear presentation of code and the non-linear process of exploratory data analysis. We present StickyLand, a notebook extension for empowering users to freely organize their code in non-linear ways. With sticky cells that are always shown on the screen, users can quickly access their notes, instantly observe experiment results, and easily build interactive dashboards that support complex visual analytics. Case studies highlight how our tool can enhance notebook users's productivity and identify opportunities for future notebook designs. StickyLand is available at https://github.com/xiaohk/stickyland.

Beyond scaling base models with more data or parameters, fine-tuned adapters provide an alternative way to generate high fidelity, custom images at reduced costs. As such, adapters have been widely adopted by open-source communities, accumulating a database of over 100K adapters-most of which are highly customized with insufficient descriptions. This paper explores the problem of matching the prompt to a set of relevant adapters, built on recent work that highlight the performance gains of composing adapters. We introduce Stylus, which efficiently selects and automatically composes task-specific adapters based on a prompt's keywords. Stylus outlines a three-stage approach that first summarizes adapters with improved descriptions and embeddings, retrieves relevant adapters, and then further assembles adapters based on prompts' keywords by checking how well they fit the prompt. To evaluate Stylus, we developed StylusDocs, a curated dataset featuring 75K adapters with pre-computed adapter embeddings. In our evaluation on popular Stable Diffusion checkpoints, Stylus achieves greater CLIP-FID Pareto efficiency and is twice as preferred, with humans and multimodal models as evaluators, over the base model. See stylus-diffusion.github.io for more.

With the spread of the use of Text2Img diffusion models such as DALL-E 2, Imagen, Mid Journey and Stable Diffusion, one challenge that artists face is selecting the right prompts to achieve the desired artistic output. We present techniques for measuring the effect that specific words and phrases in prompts have, and (in the Appendix) present guidance on the selection of prompts to produce desired effects.

Code Large Language Models (Code LLMs) have excelled at tasks like code completion but often miss deeper semantics such as execution effects and dynamic states. This paper aims to bridge the gap between Code LLMs' reliance on static text data and the need for semantic understanding for complex tasks like debugging and program repair. We introduce a novel strategy, monologue reasoning, to train Code LLMs to reason comprehensive semantics, encompassing high-level functional descriptions, local execution effects of individual statements, and overall input/output behavior, thereby linking static code text with dynamic execution states. We begin by collecting PyX, a clean Python corpus of fully executable code samples with functional descriptions and test cases. We propose training Code LLMs not only to write code but also to understand code semantics by reasoning about key properties, constraints, and execution behaviors using natural language, mimicking human verbal debugging, i.e., rubber-duck debugging. This approach led to the development of SemCoder, a Code LLM with only 6.7B parameters, which shows competitive performance with GPT-3.5-turbo on code generation and execution reasoning tasks. SemCoder achieves 79.3% on HumanEval (GPT-3.5-turbo: 76.8%), 63.6% on CRUXEval-I (GPT-3.5-turbo: 50.3%), and 63.9% on CRUXEval-O (GPT-3.5-turbo: 59.0%). We also study the effectiveness of SemCoder's monologue-style execution reasoning compared to concrete scratchpad reasoning, showing that our approach integrates semantics from multiple dimensions more smoothly. Finally, we demonstrate the potential of applying learned semantics to improve Code LLMs' debugging and self-refining capabilities. Our data, code, and models are available at: https://github.com/ARiSE-Lab/SemCoder.

Graphical User Interface (GUI) automation holds significant promise for assisting users with complex tasks, thereby boosting human productivity. Existing works leveraging Large Language Model (LLM) or LLM-based AI agents have shown capabilities in automating tasks on Android and Web platforms. However, these tasks are primarily aimed at simple device usage and entertainment operations. This paper presents a novel benchmark, AssistGUI, to evaluate whether models are capable of manipulating the mouse and keyboard on the Windows platform in response to user-requested tasks. We carefully collected a set of 100 tasks from nine widely-used software applications, such as, After Effects and MS Word, each accompanied by the necessary project files for better evaluation. Moreover, we propose an advanced Actor-Critic Embodied Agent framework, which incorporates a sophisticated GUI parser driven by an LLM-agent and an enhanced reasoning mechanism adept at handling lengthy procedural tasks. Our experimental results reveal that our GUI Parser and Reasoning mechanism outshine existing methods in performance. Nevertheless, the potential remains substantial, with the best model attaining only a 46% success rate on our benchmark. We conclude with a thorough analysis of the current methods' limitations, setting the stage for future breakthroughs in this domain.

Large language models often default to step-by-step computation even when efficient numerical shortcuts are available. This raises a basic question: do they exhibit number sense in a human-like behavioral sense, i.e., the ability to recognize numerical structure, apply shortcuts when appropriate, and avoid them when they are not? We introduce SenseMath, a controlled benchmark for evaluating structure-sensitive numerical reasoning in LLMs. SenseMath contains 4,800 items spanning eight shortcut categories and four digit scales, with matched strong-shortcut, weak-shortcut, and control variants. It supports three evaluation settings of increasing cognitive demand: Shortcut Use (whether models can apply shortcuts on shortcut-amenable problems); Applicability Judgment (whether they can recognize when a shortcut is appropriate or misleading); and Problem Generation (whether they can generate new problem items that correctly admit a given type of shortcut). Our evaluation across five LLMs, ranging from GPT-4o-mini to Llama-3.1-8B, shows a consistent pattern: when explicitly prompted, models readily adopt shortcut strategies and achieve substantial accuracy gains on shortcut-amenable items (up to 15%), yet under standard chain-of-thought prompting they spontaneously employ such strategies in fewer than 40% of cases, even when they demonstrably possess the requisite capability. Moreover, this competence is confined to the Use level; models systematically over-generalise shortcuts to problems where they do not apply, and fail to generate valid shortcut-bearing problems from scratch. Together, these results suggest that current LLMs exhibit procedural shortcut fluency without the structural understanding of when and why shortcuts work that underlies human number sense.

Large language models (LLMs) have demonstrated remarkable performance across various real-world tasks. However, they often struggle to fully comprehend and effectively utilize their input contexts, resulting in responses that are unfaithful or hallucinated. This difficulty increases for contexts that are long or contain distracting information, which can divert LLMs from fully capturing essential evidence. To address this issue, many works use prompting to help LLMs utilize contextual information more faithfully. For instance, iterative prompting highlights key information in two steps that first ask the LLM to identify important pieces of context and then derive answers accordingly. However, prompting methods are constrained to highlighting key information implicitly in token space, which is often insufficient to fully steer the model's attention. To improve model faithfulness more reliably, we propose AutoPASTA, a method that automatically identifies key contextual information and explicitly highlights it by steering an LLM's attention scores. Like prompting, AutoPASTA is applied at inference time and does not require changing any model parameters. Our experiments on open-book QA demonstrate that AutoPASTA effectively enables models to grasp essential contextual information, leading to substantially improved model faithfulness and performance, e.g., an average improvement of 7.95% for LLAMA3-70B-Instruct. Code will be publicly available at https://github.com/QingruZhang/AutoPASTA .

Skill-distillation pipelines learn reusable rules from LLM agent trajectories, but they lack a key signal: how much each step costs. Without per-step cost, a pipeline cannot distinguish adding a missing step to fix a bug from removing an expensive step that never affected the outcome. We introduce ClawTrace, an agent tracing platform that records every LLM call, tool use, and sub-agent spawn during an agent session and compiles each session into a TraceCard: a compact YAML summary with per-step USD cost, token counts, and redundancy flags. Built on ClawTrace, CostCraft is a distillation pipeline that reads TraceCards and produces three types of skill patches. Preserve patches keep behaviors that led to success. Prune patches remove expensive steps that did not matter, each backed by a counterfactual argument against a named high-cost step. Repair patches fix failures grounded in oracle evidence. Ablations on 30 held-out SpreadsheetBench tasks show that both cost attribution and prune patches independently reduce quality regressions. When the same skill is applied to 30 unrelated SkillsBench tasks, an unexpected asymmetry emerges: prune rules transferred across benchmarks and cut median cost by 32%, while preserve rules, trained on benchmark-specific conventions, caused regressions on new task types. We release ClawTrace and TraceCards as open infrastructure for cost-aware agent research.

The abilities of long-context language models (LMs) are often evaluated using the "Needle-in-a-Haystack" (NIAH) test, which comprises tasks designed to assess a model's ability to identify specific information ("needle") within large text sequences ("haystack"). While these benchmarks measure how well models understand long-context input sequences, they do not effectively gauge the quality of long-form text generation--a critical aspect for applications such as design proposals and creative writing. To address this gap, we have introduced a new long-form text evaluation benchmark, Spinning the Golden Thread (SGT), which tests models' ability to identify specific events within generated long text sequences. In this benchmark, we prompt long-context LMs to create long-form text that must include particular events or constraints and evaluate their ability to incorporate these elements. We evaluated ten long-context LMs across four distinct scenarios, three types of prompt instructions, and two different generation-length settings (16K and 32K). Although these models perform well on NIAH benchmarks, none demonstrated satisfactory performance on the Spinning the Golden Thread, raising concerns about their ability to generate coherent long-form text that follows instructions. Additionally, as the length of the generated text increases, all models exhibit a significant drop in performance.

Current foundation models exhibit impressive capabilities when prompted either with text only or with both image and text inputs. But do their capabilities change depending on the input modality? In this work, we propose IsoBench, a benchmark dataset containing problems from four major areas: math, science, algorithms, and games. Each example is presented with multiple isomorphic representations of inputs, such as visual, textual, and mathematical presentations. IsoBench provides fine-grained feedback to diagnose performance gaps caused by the form of the representation. Across various foundation models, we observe that on the same problem, models have a consistent preference towards textual representations. Most prominently, when evaluated on all IsoBench problems, Claude-3 Opus performs 28.7 points worse when provided with images instead of text; similarly, GPT-4 Turbo is 18.7 points worse and Gemini Pro is 14.9 points worse. Finally, we present two prompting techniques, IsoCombination and IsoScratchPad, which improve model performance by considering combinations of, and translations between, different input representations.

This paper, for the very first time, introduces human sketches to the landscape of XAI (Explainable Artificial Intelligence). We argue that sketch as a ``human-centred'' data form, represents a natural interface to study explainability. We focus on cultivating sketch-specific explainability designs. This starts by identifying strokes as a unique building block that offers a degree of flexibility in object construction and manipulation impossible in photos. Following this, we design a simple explainability-friendly sketch encoder that accommodates the intrinsic properties of strokes: shape, location, and order. We then move on to define the first ever XAI task for sketch, that of stroke location inversion SLI. Just as we have heat maps for photos, and correlation matrices for text, SLI offers an explainability angle to sketch in terms of asking a network how well it can recover stroke locations of an unseen sketch. We offer qualitative results for readers to interpret as snapshots of the SLI process in the paper, and as GIFs on the project page. A minor but interesting note is that thanks to its sketch-specific design, our sketch encoder also yields the best sketch recognition accuracy to date while having the smallest number of parameters. The code is available at https://sketchxai.github.io.

Current text-to-image generation models often struggle to follow textual instructions, especially the ones requiring spatial reasoning. On the other hand, Large Language Models (LLMs), such as GPT-4, have shown remarkable precision in generating code snippets for sketching out text inputs graphically, e.g., via TikZ. In this work, we introduce Control-GPT to guide the diffusion-based text-to-image pipelines with programmatic sketches generated by GPT-4, enhancing their abilities for instruction following. Control-GPT works by querying GPT-4 to write TikZ code, and the generated sketches are used as references alongside the text instructions for diffusion models (e.g., ControlNet) to generate photo-realistic images. One major challenge to training our pipeline is the lack of a dataset containing aligned text, images, and sketches. We address the issue by converting instance masks in existing datasets into polygons to mimic the sketches used at test time. As a result, Control-GPT greatly boosts the controllability of image generation. It establishes a new state-of-art on the spatial arrangement and object positioning generation and enhances users' control of object positions, sizes, etc., nearly doubling the accuracy of prior models. Our work, as a first attempt, shows the potential for employing LLMs to enhance the performance in computer vision tasks.

Keypoint detection, integral to modern machine perception, faces challenges in few-shot learning, particularly when source data from the same distribution as the query is unavailable. This gap is addressed by leveraging sketches, a popular form of human expression, providing a source-free alternative. However, challenges arise in mastering cross-modal embeddings and handling user-specific sketch styles. Our proposed framework overcomes these hurdles with a prototypical setup, combined with a grid-based locator and prototypical domain adaptation. We also demonstrate success in few-shot convergence across novel keypoints and classes through extensive experiments.

Large language models (LLMs) have demonstrated the promise to revolutionize the field of software engineering. Among other things, LLM agents are rapidly gaining momentum in their application to software development, with practitioners claiming a multifold productivity increase after adoption. Yet, empirical evidence is lacking around these claims. In this paper, we estimate the causal effect of adopting a widely popular LLM agent assistant, namely Cursor, on development velocity and software quality. The estimation is enabled by a state-of-the-art difference-in-differences design comparing Cursor-adopting GitHub projects with a matched control group of similar GitHub projects that do not use Cursor. We find that the adoption of Cursor leads to a significant, large, but transient increase in project-level development velocity, along with a significant and persistent increase in static analysis warnings and code complexity. Further panel generalized method of moments estimation reveals that the increase in static analysis warnings and code complexity acts as a major factor causing long-term velocity slowdown. Our study carries implications for software engineering practitioners, LLM agent assistant designers, and researchers.

Natural language and images are commonly used as goal representations in goal-conditioned imitation learning (IL). However, natural language can be ambiguous and images can be over-specified. In this work, we propose hand-drawn sketches as a modality for goal specification in visual imitation learning. Sketches are easy for users to provide on the fly like language, but similar to images they can also help a downstream policy to be spatially-aware and even go beyond images to disambiguate task-relevant from task-irrelevant objects. We present RT-Sketch, a goal-conditioned policy for manipulation that takes a hand-drawn sketch of the desired scene as input, and outputs actions. We train RT-Sketch on a dataset of paired trajectories and corresponding synthetically generated goal sketches. We evaluate this approach on six manipulation skills involving tabletop object rearrangements on an articulated countertop. Experimentally we find that RT-Sketch is able to perform on a similar level to image or language-conditioned agents in straightforward settings, while achieving greater robustness when language goals are ambiguous or visual distractors are present. Additionally, we show that RT-Sketch has the capacity to interpret and act upon sketches with varied levels of specificity, ranging from minimal line drawings to detailed, colored drawings. For supplementary material and videos, please refer to our website: http://rt-sketch.github.io.

Large language models (LLMs) are shaping a new user interface (UI) paradigm in writing tools by enabling users to generate text through prompts. This paradigm shifts some creative control from the user to the system, thereby diminishing the user's authorship and autonomy in the writing process. To restore autonomy, we introduce Textfocals, a UI prototype designed to investigate a human-centered approach that emphasizes the user's role in writing. Textfocals supports the writing process by providing LLM-generated summaries, questions, and advice (i.e., LLM views) in a sidebar of a text editor, encouraging reflection and self-driven revision in writing without direct text generation. Textfocals' UI affordances, including contextually adaptive views and scaffolding for prompt selection and customization, offer a novel way to interact with LLMs where users maintain full authorship of their writing. A formative user study with Textfocals showed promising evidence that this approach might help users develop underdeveloped ideas, cater to the rhetorical audience, and clarify their writing. However, the study also showed interaction design challenges related to document navigation and scoping, prompt engineering, and context management. Our work highlights the breadth of the design space of writing support interfaces powered by generative AI that maintain authorship integrity.

Crafting effective prompts for code generation or editing with Large Language Models (LLMs) is not an easy task. Particularly, the absence of immediate, stable feedback during prompt crafting hinders effective interaction, as users are left to mentally imagine possible outcomes until the code is generated. In response, we introduce Language-Oriented Code Sketching, an interactive approach that provides instant, incremental feedback in the form of code sketches (i.e., incomplete code outlines) during prompt crafting. This approach converts a prompt into a code sketch by leveraging the inherent linguistic structures within the prompt and applying classic natural language processing techniques. The sketch then serves as an intermediate placeholder that not only previews the intended code structure but also guides the LLM towards the desired code, thereby enhancing human-LLM interaction. We conclude by discussing the approach's applicability and future plans.

Following the rise of large language models (LLMs), many studies have emerged in recent years focusing on exploring the adoption of LLM-based tools for software development by novice developers: computer science/software engineering students and early-career industry developers with two years or less of professional experience. These studies have sought to understand the perspectives of novice developers on using these tools, a critical aspect of the successful adoption of LLMs in software engineering. To systematically collect and summarise these studies, we conducted a systematic literature review (SLR) following the guidelines by Kitchenham et al. on 80 primary studies published between April 2022 and June 2025 to answer four research questions (RQs). In answering RQ1, we categorised the study motivations and methodological approaches. In RQ2, we identified the software development tasks for which novice developers use LLMs. In RQ3, we categorised the advantages, challenges, and recommendations discussed in the studies. Finally, we discuss the study limitations and future research needs suggested in the primary studies in answering RQ4. Throughout the paper, we also indicate directions for future work and implications for software engineering researchers, educators, and developers. Our research artifacts are publicly available at https://github.com/Samuellucas97/SupplementaryInfoPackage-SLR.

Sketches are a natural and accessible medium for UI designers to conceptualize early-stage ideas. However, existing research on UI/UX automation often requires high-fidelity inputs like Figma designs or detailed screenshots, limiting accessibility and impeding efficient design iteration. To bridge this gap, we introduce Sketch2Code, a benchmark that evaluates state-of-the-art Vision Language Models (VLMs) on automating the conversion of rudimentary sketches into webpage prototypes. Beyond end-to-end benchmarking, Sketch2Code supports interactive agent evaluation that mimics real-world design workflows, where a VLM-based agent iteratively refines its generations by communicating with a simulated user, either passively receiving feedback instructions or proactively asking clarification questions. We comprehensively analyze ten commercial and open-source models, showing that Sketch2Code is challenging for existing VLMs; even the most capable models struggle to accurately interpret sketches and formulate effective questions that lead to steady improvement. Nevertheless, a user study with UI/UX experts reveals a significant preference for proactive question-asking over passive feedback reception, highlighting the need to develop more effective paradigms for multi-turn conversational agents.

Many recent prompting strategies for large language models (LLMs) query the model multiple times sequentially -- first to produce intermediate results and then the final answer. However, using these methods, both decoder and model are unaware of potential follow-up prompts, leading to disconnected and undesirably wordy intermediate responses. In this work, we address this issue by proposing prompt sketching, a new prompting paradigm in which an LLM does not only respond by completing a prompt, but by predicting values for multiple variables in a template. This way, sketching grants users more control over the generation process, e.g., by providing a reasoning framework via intermediate instructions, leading to better overall results. The key idea enabling sketching with existing, autoregressive models is to adapt the decoding procedure to also score follow-up instructions during text generation, thus optimizing overall template likelihood in inference. Our experiments show that in a zero-shot setting, prompt sketching outperforms existing, sequential prompting schemes such as direct asking or chain-of-thought on 7 out of 8 LLM benchmarking tasks, including state tracking, arithmetic reasoning, and general question answering. To facilitate future use, we release a number of generic, yet effective sketches applicable to many tasks, and an open source library called dclib, powering our sketch-aware decoders.

AI agents may be able to automate your inbox, but can they automate other routine aspects of your life? Everyday online tasks offer a realistic yet unsolved testbed for evaluating the next generation of AI agents. To this end, we introduce ClawBench, an evaluation framework of 153 simple tasks that people need to accomplish regularly in their lives and work, spanning 144 live platforms across 15 categories, from completing purchases and booking appointments to submitting job applications. These tasks require demanding capabilities beyond existing benchmarks, such as obtaining relevant information from user-provided documents, navigating multi-step workflows across diverse platforms, and write-heavy operations like filling in many detailed forms correctly. Unlike existing benchmarks that evaluate agents in offline sandboxes with static pages, ClawBench operates on production websites, preserving the full complexity, dynamic nature, and challenges of real-world web interaction. A lightweight interception layer captures and blocks only the final submission request, ensuring safe evaluation without real-world side effects. Our evaluations of 7 frontier models show that both proprietary and open-source models can complete only a small portion of these tasks. For example, Claude Sonnet 4.6 achieves only 33.3%. Progress on ClawBench brings us closer to AI agents that can function as reliable general-purpose assistants.

Code LLMs are being rapidly deployed and there is evidence that they can make professional programmers more productive. Current benchmarks for code generation measure whether models generate correct programs given an expert prompt. In this paper, we present a new benchmark containing multiple prompts per problem, written by a specific population of non-expert prompters: beginning programmers. StudentEval contains 1,749 prompts for 48 problems, written by 80 students who have only completed one semester of Python programming. Our students wrote these prompts while working interactively with a Code LLM, and we observed very mixed success rates. We use StudentEval to evaluate 5 Code LLMs and find that StudentEval is a better discriminator of model performance than existing benchmarks. We analyze the prompts and find significant variation in students' prompting techniques. We also find that nondeterministic LLM sampling could mislead students into thinking that their prompts are more (or less) effective than they actually are, which has implications for how to teach with Code LLMs.

Current sketch extraction methods either require extensive training or fail to capture a wide range of artistic styles, limiting their practical applicability and versatility. We introduce Mixture-of-Self-Attention (MixSA), a training-free sketch extraction method that leverages strong diffusion priors for enhanced sketch perception. At its core, MixSA employs a mixture-of-self-attention technique, which manipulates self-attention layers by substituting the keys and values with those from reference sketches. This allows for the seamless integration of brushstroke elements into initial outline images, offering precise control over texture density and enabling interpolation between styles to create novel, unseen styles. By aligning brushstroke styles with the texture and contours of colored images, particularly in late decoder layers handling local textures, MixSA addresses the common issue of color averaging by adjusting initial outlines. Evaluated with various perceptual metrics, MixSA demonstrates superior performance in sketch quality, flexibility, and applicability. This approach not only overcomes the limitations of existing methods but also empowers users to generate diverse, high-fidelity sketches that more accurately reflect a wide range of artistic expressions.

Deep learning and language models are increasingly dominating automated program repair research. While previous generate-and-validate approaches were able to find and use fix ingredients on a file or even project level, neural language models are limited to the code that fits their input window. In this work we investigate how important identifier ingredients are in neural program repair and present ScanFix, an approach that leverages an additional scanner model to extract identifiers from a bug's file and potentially project-level context. We find that lack of knowledge of far-away identifiers is an important cause of failed repairs. Augmenting repair model input with scanner-extracted identifiers yields relative improvements of up to 31%. However, ScanFix is outperformed by a model with a large input window (> 5k tokens). When passing ingredients from the ground-truth fix, improvements are even higher. This shows that, with refined extraction techniques, ingredient scanning, similar to fix candidate ranking, could have the potential to become an important subtask of future automated repair systems. At the same time, it also demonstrates that this idea is subject to Sutton's bitter lesson and may be rendered unnecessary by new code models with ever-increasing context windows.

Writing is a cognitively demanding task involving continuous decision-making, heavy use of working memory, and frequent switching between multiple activities. Scholarly writing is particularly complex as it requires authors to coordinate many pieces of multiform knowledge. To fully understand writers' cognitive thought process, one should fully decode the end-to-end writing data (from individual ideas to final manuscript) and understand their complex cognitive mechanisms in scholarly writing. We introduce ScholaWrite dataset, the first-of-its-kind keystroke logs of an end-to-end scholarly writing process for complete manuscripts, with thorough annotations of cognitive writing intentions behind each keystroke. Our dataset includes LaTeX-based keystroke data from five preprints with nearly 62K total text changes and annotations across 4 months of paper writing. ScholaWrite shows promising usability and applications (e.g., iterative self-writing) for the future development of AI writing assistants for academic research, which necessitate complex methods beyond LLM prompting. Our experiments clearly demonstrated the importance of collection of end-to-end writing data, rather than the final manuscript, for the development of future writing assistants to support the cognitive thinking process of scientists. Our de-identified dataset, demo, and code repository are available on our project page.

While paper instructions are one of the mainstream medium for sharing knowledge, consuming such instructions and translating them into activities are inefficient due to the lack of connectivity with physical environment. We present PaperToPlace, a novel workflow comprising an authoring pipeline, which allows the authors to rapidly transform and spatialize existing paper instructions into MR experience, and a consumption pipeline, which computationally place each instruction step at an optimal location that is easy to read and do not occlude key interaction areas. Our evaluations of the authoring pipeline with 12 participants demonstrated the usability of our workflow and the effectiveness of using a machine learning based approach to help extracting the spatial locations associated with each steps. A second within-subject study with another 12 participants demonstrates the merits of our consumption pipeline by reducing efforts of context switching, delivering the segmented instruction steps and offering the hands-free affordances.

We present new data and semantic parsing methods for the problem of mapping English sentences to Bash commands (NL2Bash). Our long-term goal is to enable any user to perform operations such as file manipulation, search, and application-specific scripting by simply stating their goals in English. We take a first step in this domain, by providing a new dataset of challenging but commonly used Bash commands and expert-written English descriptions, along with baseline methods to establish performance levels on this task.

Today, large language models (LLMs) are taught to use new tools by providing a few demonstrations of the tool's usage. Unfortunately, demonstrations are hard to acquire, and can result in undesirable biased usage if the wrong demonstration is chosen. Even in the rare scenario that demonstrations are readily available, there is no principled selection protocol to determine how many and which ones to provide. As tasks grow more complex, the selection search grows combinatorially and invariably becomes intractable. Our work provides an alternative to demonstrations: tool documentation. We advocate the use of tool documentation, descriptions for the individual tool usage, over demonstrations. We substantiate our claim through three main empirical findings on 6 tasks across both vision and language modalities. First, on existing benchmarks, zero-shot prompts with only tool documentation are sufficient for eliciting proper tool usage, achieving performance on par with few-shot prompts. Second, on a newly collected realistic tool-use dataset with hundreds of available tool APIs, we show that tool documentation is significantly more valuable than demonstrations, with zero-shot documentation significantly outperforming few-shot without documentation. Third, we highlight the benefits of tool documentations by tackling image generation and video tracking using just-released unseen state-of-the-art models as tools. Finally, we highlight the possibility of using tool documentation to automatically enable new applications: by using nothing more than the documentation of GroundingDino, Stable Diffusion, XMem, and SAM, LLMs can re-invent the functionalities of the just-released Grounded-SAM and Track Anything models.

Although LLMs are increasing the productivity of professional programmers, existing work shows that beginners struggle to prompt LLMs to solve text-to-code tasks. Why is this the case? This paper explores two competing hypotheses about the cause of student-LLM miscommunication: (1) students simply lack the technical vocabulary needed to write good prompts, and (2) students do not understand the extent of information that LLMs need to solve code generation tasks. We study (1) with a causal intervention experiment on technical vocabulary and (2) by analyzing graphs that abstract how students edit prompts and the different failures that they encounter. We find that substance beats style: a poor grasp of technical vocabulary is merely correlated with prompt failure; that the information content of prompts predicts success; that students get stuck making trivial edits; and more. Our findings have implications for the use of LLMs in programming education, and for efforts to make computing more accessible with LLMs.

Recent theoretical work has identified surprisingly simple reasoning problems, such as checking if two nodes in a graph are connected or simulating finite-state machines, that are provably unsolvable by standard transformers that answer immediately after reading their input. However, in practice, transformers' reasoning can be improved by allowing them to use a "chain of thought" or "scratchpad", i.e., generate and condition on a sequence of intermediate tokens before answering. Motivated by this, we ask: Does such intermediate generation fundamentally extend the computational power of a decoder-only transformer? We show that the answer is yes, but the amount of increase depends crucially on the amount of intermediate generation. For instance, we find that transformer decoders with a logarithmic number of decoding steps (w.r.t. the input length) push the limits of standard transformers only slightly, while a linear number of decoding steps, assuming a slight generalization to standard pre-norm, adds a clear new ability (under standard complexity conjectures): recognizing all regular languages. Our results also imply that linear steps keep transformer decoders within context-sensitive languages, and polynomial steps with generalized pre-norm make them recognize exactly the class of polynomial-time solvable problems -- the first exact characterization of a type of transformers in terms of standard complexity classes. Together, our results provide a nuanced framework for understanding how the length of a transformer's chain of thought or scratchpad impacts its reasoning power.

ResearchPilot is an open-source, self-hostable multi-agent system for literature-review assistance. Given a natural-language research question, it retrieves papers from Semantic Scholar and arXiv, extracts structured findings from paper abstracts, synthesizes cross-paper patterns, and drafts a citation-aware related-work section. The system combines FastAPI, Next.js, DSPy, SQLite, and Qdrant in a local-first architecture that supports bring-your-own-key model access and remote-or-local embeddings. This paper describes the system design, typed agent interfaces, persistence and history-search mechanisms, and the engineering tradeoffs involved in building a transparent research assistant. Rather than claiming algorithmic novelty, we present ResearchPilot as a systems contribution and evaluate it through automated tests and end-to-end local runs. We discuss limitations including external API rate limits, abstract-only extraction, incomplete corpus coverage, and the lack of citation verification.

Text entry is an essential task in our day-to-day digital interactions. Numerous intelligent features have been developed to streamline this process, making text entry more effective, efficient, and fluid. These improvements include sentence prediction and user personalization. However, as deep learning-based language models become the norm for these advanced features, the necessity for data collection and model fine-tuning increases. These challenges can be mitigated by harnessing the in-context learning capability of large language models such as GPT-3.5. This unique feature allows the language model to acquire new skills through prompts, eliminating the need for data collection and fine-tuning. Consequently, large language models can learn various text prediction techniques. We initially showed that, for a sentence prediction task, merely prompting GPT-3.5 surpassed a GPT-2 backed system and is comparable with a fine-tuned GPT-3.5 model, with the latter two methods requiring costly data collection, fine-tuning and post-processing. However, the task of prompting large language models to specialize in specific text prediction tasks can be challenging, particularly for designers without expertise in prompt engineering. To address this, we introduce Promptor, a conversational prompt generation agent designed to engage proactively with designers. Promptor can automatically generate complex prompts tailored to meet specific needs, thus offering a solution to this challenge. We conducted a user study involving 24 participants creating prompts for three intelligent text entry tasks, half of the participants used Promptor while the other half designed prompts themselves. The results show that Promptor-designed prompts result in a 35% increase in similarity and 22% in coherence over those by designers.

The web is full of guidance on a wide variety of tasks, from changing the oil in your car to baking an apple pie. However, as content is created independently, a single task could have thousands of corresponding procedural texts. This makes it difficult for users to view the bigger picture and understand the multiple ways the task could be accomplished. In this work we propose an unsupervised learning approach for summarizing multiple procedural texts into an intuitive graph representation, allowing users to easily explore commonalities and differences. We demonstrate our approach on recipes, a prominent example of procedural texts. User studies show that our representation is intuitive and coherent and that it has the potential to help users with several sensemaking tasks, including adapting recipes for a novice cook and finding creative ways to spice up a dish.

Optimizing the phrasing of argumentative text is crucial in higher education and professional development. However, assessing whether and how the different claims in a text should be revised is a hard task, especially for novice writers. In this work, we explore the main challenges to identifying argumentative claims in need of specific revisions. By learning from collaborative editing behaviors in online debates, we seek to capture implicit revision patterns in order to develop approaches aimed at guiding writers in how to further improve their arguments. We systematically compare the ability of common word embedding models to capture the differences between different versions of the same text, and we analyze their impact on various types of writing issues. To deal with the noisy nature of revision-based corpora, we propose a new sampling strategy based on revision distance. Opposed to approaches from prior work, such sampling can be done without employing additional annotations and judgments. Moreover, we provide evidence that using contextual information and domain knowledge can further improve prediction results. How useful a certain type of context is, depends on the issue the claim is suffering from, though.

The paper investigates using a Large Language Model (LLM) to automatically perform web software tasks using click, scroll, and text input operations. Previous approaches, such as reinforcement learning (RL) or imitation learning, are inefficient to train and task-specific. Our method uses filtered Document Object Model (DOM) elements as observations and performs tasks step-by-step, sequentially generating small programs based on the current observations. We use in-context learning, either benefiting from a single manually provided example, or an automatically generated example based on a successful zero-shot trial. We evaluate the proposed method on the MiniWob++ benchmark. With only one in-context example, our WebWISE method achieves similar or better performance than other methods that require many demonstrations or trials.

Sketch editing is central to digital illustration, yet existing image editing systems struggle to preserve the sparse, style-sensitive structure of line art while supporting both high-level semantic changes and precise local redrawing. We present SketchAssist, an interactive sketch drawing assistant that accelerates creation by unifying instruction-guided global edits with line-guided region redrawing, while keeping unrelated regions and overall composition intact. To enable this assistant at scale, we introduce a controllable data generation pipeline that (i) constructs attribute-addition sequences from attribute-free base sketches, (ii) forms multi-step edit chains via cross-sequence sampling, and (iii) expands stylistic coverage with a style-preserving attribute-removal model applied to diverse sketches. Building on this data, SketchAssist employs a unified sketch editing framework with minimal changes to DiT-based editors. We repurpose the RGB channels to encode the inputs, enabling seamless switching between instruction-guided edits and line-guided redrawing within a single input interface. To further specialize behavior across modes, we integrate a task-guided mixture-of-experts into LoRA layers, routing by text and visual cues to improve semantic controllability, structural fidelity, and style preservation. Extensive experiments show state-of-the-art results on both tasks, with superior instruction adherence and style/structure preservation compared to recent baselines. Together, our dataset and SketchAssist provide a practical, controllable assistant for sketch creation and revision.

State-of-the-art visual generative AI tools hold immense potential to assist users in the early ideation stages of creative tasks -- offering the ability to generate (rather than search for) novel and unprecedented (instead of existing) images of considerable quality that also adhere to boundless combinations of user specifications. However, many large-scale text-to-image systems are designed for broad applicability, yielding conventional output that may limit creative exploration. They also employ interaction methods that may be difficult for beginners. Given that creative end users often operate in diverse, context-specific ways that are often unpredictable, more variation and personalization are necessary. We introduce POET, a real-time interactive tool that (1) automatically discovers dimensions of homogeneity in text-to-image generative models, (2) expands these dimensions to diversify the output space of generated images, and (3) learns from user feedback to personalize expansions. An evaluation with 28 users spanning four creative task domains demonstrated POET's ability to generate results with higher perceived diversity and help users reach satisfaction in fewer prompts during creative tasks, thereby prompting them to deliberate and reflect more on a wider range of possible produced results during the co-creative process. Focusing on visual creativity, POET offers a first glimpse of how interaction techniques of future text-to-image generation tools may support and align with more pluralistic values and the needs of end users during the ideation stages of their work.

Graphical user interface (GUI) grounding, the ability to map natural language instructions to specific actions on graphical user interfaces, remains a critical bottleneck in computer use agent development. Current benchmarks oversimplify grounding tasks as short referring expressions, failing to capture the complexity of real-world interactions that require software commonsense, layout understanding, and fine-grained manipulation capabilities. To address these limitations, we introduce OSWorld-G, a comprehensive benchmark comprising 564 finely annotated samples across diverse task types including text matching, element recognition, layout understanding, and precise manipulation. Additionally, we synthesize and release the largest computer use grounding dataset Jedi, which contains 4 million examples through multi-perspective decoupling of tasks. Our multi-scale models trained on Jedi demonstrate its effectiveness by outperforming existing approaches on ScreenSpot-v2, ScreenSpot-Pro, and our OSWorld-G. Furthermore, we demonstrate that improved grounding with Jedi directly enhances agentic capabilities of general foundation models on complex computer tasks, improving from 5% to 27% on OSWorld. Through detailed ablation studies, we identify key factors contributing to grounding performance and verify that combining specialized data for different interface elements enables compositional generalization to novel interfaces. All benchmark, data, checkpoints, and code are open-sourced and available at https://osworld-grounding.github.io.

As academic paper publication numbers grow exponentially, conducting in-depth surveys with LLMs automatically has become an inevitable trend. Outline writing, which aims to systematically organize related works, is critical for automated survey generation. Yet existing automatic survey methods treat outline writing as mere workflow steps in the overall pipeline. Such template-based workflows produce outlines that lack in-depth understanding of the survey topic and fine-grained styles. To address these limitations, we propose Meow, the first metadata-driven outline writing framework that produces organized and faithful outlines efficiently. Specifically, we first formulate outline writing as an end-to-end task that generates hierarchical structured outlines from paper metadata. We then curate a high-quality dataset of surveys from arXiv, bioRxiv, and medRxiv, and establish systematic evaluation metrics for outline quality assessment. Finally, we employ a two-stage training approach combining supervised fine-tuning and reinforcement learning. Our 8B reasoning model demonstrates strong performance with high structural fidelity and stylistic coherence.

Recent progress in Large Language Models (LLMs) has produced models that exhibit remarkable performance across a variety of NLP tasks. However, it remains unclear whether the existing focus of NLP research accurately captures the genuine requirements of human users. This paper provides a comprehensive analysis of the divergence between current NLP research and the needs of real-world NLP applications via a large-scale collection of user-GPT conversations. We analyze a large-scale collection of real user queries to GPT. We compare these queries against existing NLP benchmark tasks and identify a significant gap between the tasks that users frequently request from LLMs and the tasks that are commonly studied in academic research. For example, we find that tasks such as ``design'' and ``planning'' are prevalent in user interactions but are largely neglected or different from traditional NLP benchmarks. We investigate these overlooked tasks, dissect the practical challenges they pose, and provide insights toward a roadmap to make LLMs better aligned with user needs.

Large language models are increasingly used as programming agents for repository level software engineering tasks. While recent benchmarks evaluate correctness in realistic codebases, they largely treat tasks as independent and do not assess whether agents can reuse experience across related problems. As a result, the ability of agents to accumulate, retrieve, and apply prior experience, as well as the efficiency gains from such reuse, remains difficult to measure. We introduce SWE-ContextBench, a benchmark designed to explicitly evaluate experience reuse in programming agents. Built on SWE-Bench Lite, SWE-ContextBench augments 300 base tasks with 99 related tasks derived from real dependency and reference relationships among GitHub issues and pull requests, forming task sequences with shared context. The benchmark evaluates agents along three complementary dimensions: prediction accuracy, time efficiency, and cost efficiency. Using SWE-ContextBench, we study multiple experience reuse settings, including oracle guided and autonomous retrieval, as well as full execution trajectories and compact summaries. Our results show that correctly selected summarized experience improves resolution accuracy and substantially reduces runtime and token cost, particularly on harder tasks. In contrast, unfiltered or incorrectly selected experience provides limited or negative benefits. These findings highlight the importance of experience representation and retrieval quality, and position SWE-ContextBench as a principled benchmark for studying experience reuse in programming agents.

This paper introduces RAISE (Reasoning and Acting through Scratchpad and Examples), an advanced architecture enhancing the integration of Large Language Models (LLMs) like GPT-4 into conversational agents. RAISE, an enhancement of the ReAct framework, incorporates a dual-component memory system, mirroring human short-term and long-term memory, to maintain context and continuity in conversations. It entails a comprehensive agent construction scenario, including phases like Conversation Selection, Scene Extraction, CoT Completion, and Scene Augmentation, leading to the LLMs Training phase. This approach appears to enhance agent controllability and adaptability in complex, multi-turn dialogues. Our preliminary evaluations in a real estate sales context suggest that RAISE has some advantages over traditional agents, indicating its potential for broader applications. This work contributes to the AI field by providing a robust framework for developing more context-aware and versatile conversational agents.

Effective evaluation of multi-hop tool use is critical for analyzing the understanding, reasoning, and function-calling capabilities of large language models (LLMs). However, progress has been hindered by a lack of reliable evaluation datasets. To address this, we present ToolHop, a dataset comprising 995 user queries and 3,912 associated tools, specifically designed for rigorous evaluation of multi-hop tool use. ToolHop ensures diverse queries, meaningful interdependencies, locally executable tools, detailed feedback, and verifiable answers through a novel query-driven data construction approach that includes tool creation, document refinement, and code generation. We evaluate 14 LLMs across five model families (i.e., LLaMA3.1, Qwen2.5, Gemini1.5, Claude3.5, and GPT), uncovering significant challenges in handling multi-hop tool-use scenarios. The leading model, GPT-4o, achieves an accuracy of 49.04%, underscoring substantial room for improvement. Further analysis reveals variations in tool-use strategies for various families, offering actionable insights to guide the development of more effective approaches. Code and data can be found in https://huggingface.co/bytedance-research/ToolHop.

Evaluation of large language models (LLMs) for code has primarily relied on static benchmarks, including HumanEval (Chen et al., 2021), which measure the ability of LLMs to generate complete code that passes unit tests. As LLMs are increasingly used as programmer assistants, we study whether gains on existing benchmarks translate to gains in programmer productivity when coding with LLMs, including time spent coding. In addition to static benchmarks, we investigate the utility of preference metrics that might be used as proxies to measure LLM helpfulness, such as code acceptance or copy rates. To do so, we introduce RealHumanEval, a web interface to measure the ability of LLMs to assist programmers, through either autocomplete or chat support. We conducted a user study (N=213) using RealHumanEval in which users interacted with six LLMs of varying base model performance. Despite static benchmarks not incorporating humans-in-the-loop, we find that improvements in benchmark performance lead to increased programmer productivity; however gaps in benchmark versus human performance are not proportional -- a trend that holds across both forms of LLM support. In contrast, we find that programmer preferences do not correlate with their actual performance, motivating the need for better, human-centric proxy signals. We also open-source RealHumanEval to enable human-centric evaluation of new models and the study data to facilitate efforts to improve code models.

Large language models (LLMs) are becoming more prevalent and have found a ubiquitous use in providing different forms of writing assistance. However, LLM-powered writing systems can frustrate users due to their limited personalization and control, which can be exacerbated when users lack experience with prompt engineering. We see design as one way to address these challenges and introduce GhostWriter, an AI-enhanced writing design probe where users can exercise enhanced agency and personalization. GhostWriter leverages LLMs to learn the user's intended writing style implicitly as they write, while allowing explicit teaching moments through manual style edits and annotations. We study 18 participants who use GhostWriter on two different writing tasks, observing that it helps users craft personalized text generations and empowers them by providing multiple ways to control the system's writing style. From this study, we present insights regarding people's relationship with AI-assisted writing and offer design recommendations for future work.

This paper, for the first time, marries large foundation models with human sketch understanding. We demonstrate what this brings -- a paradigm shift in terms of generalised sketch representation learning (e.g., classification). This generalisation happens on two fronts: (i) generalisation across unknown categories (i.e., open-set), and (ii) generalisation traversing abstraction levels (i.e., good and bad sketches), both being timely challenges that remain unsolved in the sketch literature. Our design is intuitive and centred around transferring the already stellar generalisation ability of CLIP to benefit generalised learning for sketches. We first "condition" the vanilla CLIP model by learning sketch-specific prompts using a novel auxiliary head of raster to vector sketch conversion. This importantly makes CLIP "sketch-aware". We then make CLIP acute to the inherently different sketch abstraction levels. This is achieved by learning a codebook of abstraction-specific prompt biases, a weighted combination of which facilitates the representation of sketches across abstraction levels -- low abstract edge-maps, medium abstract sketches in TU-Berlin, and highly abstract doodles in QuickDraw. Our framework surpasses popular sketch representation learning algorithms in both zero-shot and few-shot setups and in novel settings across different abstraction boundaries.

The Patent-Based Idea Generation task asks systems to turn real patents into product ideas viable within three years. We propose MK2, a prompt-centric pipeline: Gemini 2.5 drafts and iteratively edits a prompt, grafting useful fragments from weaker outputs; GPT-4.1 then uses this prompt to create one idea per patent, and an Elo loop judged by Qwen3-8B selects the best prompt-all without extra training data. Across three domains, two evaluator types, and six criteria, MK2 topped the automatic leaderboard and won 25 of 36 tests. Only the materials-chemistry track lagged, indicating the need for deeper domain grounding; yet, the results show that lightweight prompt engineering has already delivered competitive, commercially relevant ideation from patents.

We present Speakerly, a new real-time voice-based writing assistance system that helps users with text composition across various use cases such as emails, instant messages, and notes. The user can interact with the system through instructions or dictation, and the system generates a well-formatted and coherent document. We describe the system architecture and detail how we address the various challenges while building and deploying such a system at scale. More specifically, our system uses a combination of small, task-specific models as well as pre-trained language models for fast and effective text composition while supporting a variety of input modes for better usability.

Artificial Intelligence Generated Content (AIGC) has shown remarkable progress in generating realistic images. However, in this paper, we take a step "backward" and address AIGC for the most rudimentary visual modality of human sketches. Our objective is on the creative nature of sketches, and that creative sketching should take the form of an interactive process. We further enable text to drive the sketch ideation process, allowing creativity to be freely defined, while simultaneously tackling the challenge of "I can't sketch". We present a method to generate controlled sketches using a text-conditioned diffusion model trained on pixel representations of images. Our proposed approach, referred to as SketchDreamer, integrates a differentiable rasteriser of Bezier curves that optimises an initial input to distil abstract semantic knowledge from a pretrained diffusion model. We utilise Score Distillation Sampling to learn a sketch that aligns with a given caption, which importantly enable both text and sketch to interact with the ideation process. Our objective is to empower non-professional users to create sketches and, through a series of optimisation processes, transform a narrative into a storyboard by expanding the text prompt while making minor adjustments to the sketch input. Through this work, we hope to aspire the way we create visual content, democratise the creative process, and inspire further research in enhancing human creativity in AIGC. The code is available at https://github.com/WinKawaks/SketchDreamer.

Large language models are increasingly embedded into academic writing workflows, yet existing assistants remain external to the editor, preventing deep interaction with document state, structure, and revision history. This separation makes it impossible to support agentic, context-aware operations directly within LaTeX editors such as Overleaf. We present PaperDebugger, an in-editor, multi-agent, and plugin-based academic writing assistant that brings LLM-driven reasoning directly into the writing environment. Enabling such in-editor interaction is technically non-trivial: it requires reliable bidirectional synchronization with the editor, fine-grained version control and patching, secure state management, multi-agent scheduling, and extensible communication with external tools. PaperDebugger addresses these challenges through a Chrome-approved extension, a Kubernetes-native orchestration layer, and a Model Context Protocol (MCP) toolchain that integrates literature search, reference lookup, document scoring, and revision pipelines. Our demo showcases a fully integrated workflow, including localized edits, structured reviews, parallel agent execution, and diff-based updates, encapsulated within a minimal-intrusion user interface (UI). Early aggregated analytics demonstrate active user engagement and validate the practicality of an editor-native, agentic writing assistant. More details about this demo and video could be found at https://github.com/PaperDebugger/PaperDebugger.

With the goal of understanding the visual concepts that CLIP associates with text prompts, we show that the latent space of CLIP can be visualized solely in terms of linear transformations on simple geometric primitives like circles and straight lines. Although existing approaches achieve this by sketch-synthesis-through-optimization, they do so on the space of B\'ezier curves, which exhibit a wastefully large set of structures that they can evolve into, as most of them are non-essential for generating meaningful sketches. We present CLIPDrawX, an algorithm that provides significantly better visualizations for CLIP text embeddings, using only simple primitive shapes like straight lines and circles. This constrains the set of possible outputs to linear transformations on these primitives, thereby exhibiting an inherently simpler mathematical form. The synthesis process of CLIPDrawX can be tracked end-to-end, with each visual concept being explained exclusively in terms of primitives. Implementation will be released upon acceptance. Project Page: https://clipdrawx.github.io/{https://clipdrawx.github.io/}.

CodeCompose is an AI-assisted code authoring tool powered by large language models (LLMs) that provides inline suggestions to 10's of thousands of developers at Meta. In this paper, we present how we scaled the product from displaying single-line suggestions to multi-line suggestions. This evolution required us to overcome several unique challenges in improving the usability of these suggestions for developers. First, we discuss how multi-line suggestions can have a 'jarring' effect, as the LLM's suggestions constantly move around the developer's existing code, which would otherwise result in decreased productivity and satisfaction. Second, multi-line suggestions take significantly longer to generate; hence we present several innovative investments we made to reduce the perceived latency for users. These model-hosting optimizations sped up multi-line suggestion latency by 2.5x. Finally, we conduct experiments on 10's of thousands of engineers to understand how multi-line suggestions impact the user experience and contrast this with single-line suggestions. Our experiments reveal that (i) multi-line suggestions account for 42% of total characters accepted (despite only accounting for 16% for displayed suggestions) (ii) multi-line suggestions almost doubled the percentage of keystrokes saved for users from 9% to 17%. Multi-line CodeCompose has been rolled out to all engineers at Meta, and less than 1% of engineers have opted out of multi-line suggestions.

Formulating effective search queries remains a challenging task, particularly when users lack expertise in a specific domain or are not proficient in the language of the content. Providing example documents of interest might be easier for a user. However, such query-by-example scenarios are prone to concept drift, and the retrieval effectiveness is highly sensitive to the query generation method, without a clear way to incorporate user feedback. To enable exploration and to support Human-In-The-Loop experiments we propose QueryExplorer -- an interactive query generation, reformulation, and retrieval interface with support for HuggingFace generation models and PyTerrier's retrieval pipelines and datasets, and extensive logging of human feedback. To allow users to create and modify effective queries, our demo supports complementary approaches of using LLMs interactively, assisting the user with edits and feedback at multiple stages of the query formulation process. With support for recording fine-grained interactions and user annotations, QueryExplorer can serve as a valuable experimental and research platform for annotation, qualitative evaluation, and conducting Human-in-the-Loop (HITL) experiments for complex search tasks where users struggle to formulate queries.

A sketch is a distilled form of visual abstraction that conveys core concepts through simplified yet purposeful strokes while omitting extraneous detail. Despite its expressive power, quantifying the efficiency of semantic abstraction in sketches remains challenging. Existing evaluation methods that rely on reference images, low-level visual features, or recognition accuracy do not capture abstraction, the defining property of sketches. To address these limitations, we introduce SEA (Sketch Evaluation metric for Abstraction efficiency), a reference-free metric that assesses how economically a sketch represents class-defining visual elements while preserving semantic recognizability. These elements are derived per class from commonsense knowledge about features typically depicted in sketches. SEA leverages a visual question answering model to determine the presence of each element and returns a quantitative score that reflects semantic retention under visual economy. To support this metric, we present CommonSketch, the first semantically annotated sketch dataset, comprising 23,100 human-drawn sketches across 300 classes, each paired with a caption and element-level annotations. Experiments show that SEA aligns closely with human judgments and reliably discriminates levels of abstraction efficiency, while CommonSketch serves as a benchmark providing systematic evaluation of element-level sketch understanding across various vision-language models.

Memory systems are critical for role-playing agents (RPAs) to maintain long-horizon consistency. However, existing RPA memory methods (e.g., profiling) mainly rely on recurrent summarization, whose compression inevitably discards important details. To address this issue, we propose a search-based memory framework called BOOKMARKS, which actively initializes, maintains, and updates task-relevant pieces of bookmarks for the current task (e.g., character acting). A bookmark is structured as the answer to a question at a specific point in the storyline. For each current task, BOOKMARKS selects reusable existing bookmarks or initializes new ones (at storyline beginning) with useful questions. These bookmarks are then synchronized to the current story point, with their answers updated accordingly, so they can be efficiently reused in future grounding rounds. Compared with recurrent summarization, BOOKMARKS offers (1) active grounding for capturing task-specific details and (2) passive updating to avoid unnecessary computation. In implementation, BOOKMARKS supports concept, behavior, and state searches, each powered by an efficient synchronization method. BOOKMARKS significantly outperforms RPA memory baselines on 85 characters from 16 artifacts, demonstrating the effectiveness of search-based memory for RPAs.

Large language models (LLMs) have demonstrated the potential to perform high-level planning. Yet, it remains a challenge for LLMs to comprehend low-level commands, such as joint angle targets or motor torques. This paper proposes an approach to use foot contact patterns as an interface that bridges human commands in natural language and a locomotion controller that outputs these low-level commands. This results in an interactive system for quadrupedal robots that allows the users to craft diverse locomotion behaviors flexibly. We contribute an LLM prompt design, a reward function, and a method to expose the controller to the feasible distribution of contact patterns. The results are a controller capable of achieving diverse locomotion patterns that can be transferred to real robot hardware. Compared with other design choices, the proposed approach enjoys more than 50% success rate in predicting the correct contact patterns and can solve 10 more tasks out of a total of 30 tasks. Our project site is: https://saytap.github.io.

While LLMs have shown great success in understanding and generating text in traditional conversational settings, their potential for performing ill-defined complex tasks is largely under-studied. Indeed, we are yet to conduct comprehensive benchmarking studies with multiple LLMs that are exclusively focused on a complex task. However, conducting such benchmarking studies is challenging because of the large variations in LLMs' performance when different prompt types/styles are used and different degrees of detail are provided in the prompts. To address this issue, the paper proposes a general taxonomy that can be used to design prompts with specific properties in order to perform a wide range of complex tasks. This taxonomy will allow future benchmarking studies to report the specific categories of prompts used as part of the study, enabling meaningful comparisons across different studies. Also, by establishing a common standard through this taxonomy, researchers will be able to draw more accurate conclusions about LLMs' performance on a specific complex task.

Stepwise inference protocols, such as scratchpads and chain-of-thought, help language models solve complex problems by decomposing them into a sequence of simpler subproblems. Despite the significant gain in performance achieved via these protocols, the underlying mechanisms of stepwise inference have remained elusive. To address this, we propose to study autoregressive Transformer models on a synthetic task that embodies the multi-step nature of problems where stepwise inference is generally most useful. Specifically, we define a graph navigation problem wherein a model is tasked with traversing a path from a start to a goal node on the graph. Despite is simplicity, we find we can empirically reproduce and analyze several phenomena observed at scale: (i) the stepwise inference reasoning gap, the cause of which we find in the structure of the training data; (ii) a diversity-accuracy tradeoff in model generations as sampling temperature varies; (iii) a simplicity bias in the model's output; and (iv) compositional generalization and a primacy bias with in-context exemplars. Overall, our work introduces a grounded, synthetic framework for studying stepwise inference and offers mechanistic hypotheses that can lay the foundation for a deeper understanding of this phenomenon.

Code-recommendation systems, such as Copilot and CodeWhisperer, have the potential to improve programmer productivity by suggesting and auto-completing code. However, to fully realize their potential, we must understand how programmers interact with these systems and identify ways to improve that interaction. To make progress, we studied GitHub Copilot, a code-recommendation system used by millions of programmers daily. We developed CUPS, a taxonomy of common programmer activities when interacting with Copilot. Our study of 21 programmers, who completed coding tasks and retrospectively labeled their sessions with CUPS, showed that CUPS can help us understand how programmers interact with code-recommendation systems, revealing inefficiencies and time costs. Our insights reveal how programmers interact with Copilot and motivate new interface designs and metrics.

Revision behavior in adaptive writing support systems is an important and relatively new area of research that can improve the design and effectiveness of these tools, and promote students' self-regulated learning (SRL). Understanding how these tools are used is key to improving them to better support learners in their writing and learning processes. In this paper, we present a novel pipeline with insights into the revision behavior of students at scale. We leverage a data set of two groups using an adaptive writing support tool in an educational setting. With our novel pipeline, we show that the tool was effective in promoting revision among the learners. Depending on the writing feedback, we were able to analyze different strategies of learners when revising their texts, we found that users of the exemplary case improved over time and that females tend to be more efficient. Our research contributes a pipeline for measuring SRL behaviors at scale in writing tasks (i.e., engagement or revision behavior) and informs the design of future adaptive writing support systems for education, with the goal of enhancing their effectiveness in supporting student writing. The source code is available at https://github.com/lucamouchel/Understanding-Revision-Behavior.

Text-to-image generation has seen an explosion of interest since 2021. Today, beautiful and intriguing digital images and artworks can be synthesized from textual inputs ("prompts") with deep generative models. Online communities around text-to-image generation and AI generated art have quickly emerged. This paper identifies six types of prompt modifiers used by practitioners in the online community based on a 3-month ethnographic study. The novel taxonomy of prompt modifiers provides researchers a conceptual starting point for investigating the practice of text-to-image generation, but may also help practitioners of AI generated art improve their images. We further outline how prompt modifiers are applied in the practice of "prompt engineering." We discuss research opportunities of this novel creative practice in the field of Human-Computer Interaction (HCI). The paper concludes with a discussion of broader implications of prompt engineering from the perspective of Human-AI Interaction (HAI) in future applications beyond the use case of text-to-image generation and AI generated art.

Diffusion models have become widely adopted in image completion tasks, with text prompts commonly employed to ensure semantic coherence by providing high-level guidance. However, a persistent challenge arises when an object is partially obscured in the damaged region, yet its remaining parts are still visible in the background. While text prompts offer semantic direction, they often fail to precisely recover fine-grained structural details, such as the object's overall posture, ensuring alignment with the visible object information in the background. This limitation stems from the inability of text prompts to provide pixel-level specificity. To address this, we propose supplementing text-based guidance with a novel visual aid: a casual sketch, which can be roughly drawn by anyone based on visible object parts. This sketch supplies critical structural cues, enabling the generative model to produce an object structure that seamlessly integrates with the existing background. We introduce the Visual Sketch Self-Aware (VSSA) model, which integrates the casual sketch into each iterative step of the diffusion process, offering distinct advantages for partially corrupted scenarios. By blending sketch-derived features with those of the corrupted image, and leveraging text prompt guidance, the VSSA assists the diffusion model in generating images that preserve both the intended object semantics and structural consistency across the restored objects and original regions. To support this research, we created two datasets, CUB-sketch and MSCOCO-sketch, each combining images, sketches, and text. Extensive qualitative and quantitative experiments demonstrate that our approach outperforms several state-of-the-art methods.

Burrows' Delta was introduced in 2002 and has proven to be an effective tool for author attribution. Despite the fact that these are different languages, they mostly belong to the same grammatical type and use the same graphic principle to convey speech in writing: a phonemic alphabet with word separation using spaces. The question I want to address in this article is how well this attribution method works with texts in a language with a different grammatical structure and a script based on different principles. There are fewer studies analyzing the effectiveness of the Delta method on Chinese texts than on texts in European languages. I believe that such a low level of attention to Delta from sinologists is due to the structure of the scientific field dedicated to medieval Chinese poetry. Clustering based on intertextual distances worked flawlessly. Delta produced results where clustering showed that the samples of one author were most similar to each other, and Delta never confused different poets. Despite the fact that I used an unconventional approach and applied the Delta method to a language poorly suited for it, the method demonstrated its effectiveness. Tang dynasty poets are correctly identified using Delta, and the empirical pattern observed for authors writing in European standard languages has been confirmed once again.

LLM-based coding agents have shown strong performance on automated issue resolution benchmarks, yet existing evaluations largely focus on final task success, providing limited insight into how agents retrieve and use code context during problem solving. We introduce ContextBench, a process-oriented evaluation of context retrieval in coding agents. ContextBench consists of 1,136 issue-resolution tasks from 66 repositories across eight programming languages, each augmented with human-annotated gold contexts. We further implement an automated evaluation framework that tracks agent trajectories and measures context recall, precision, and efficiency throughout issue resolution. Using ContextBench, we evaluate four frontier LLMs and five coding agents. Our results show that sophisticated agent scaffolding yields only marginal gains in context retrieval ("The Bitter Lesson" of coding agents), LLMs consistently favor recall over precision, and substantial gaps exist between explored and utilized context. ContextBench augments existing end-to-end benchmarks with intermediate gold-context metrics that unbox the issue-resolution process. These contexts offer valuable intermediate signals for guiding LLM reasoning in software tasks.

This paper introduces a cost-effective robotic handwriting system designed to replicate human-like handwriting with high precision. Combining a Raspberry Pi Pico microcontroller, 3D-printed components, and a machine learning-based handwriting generation model implemented via TensorFlow, the system converts user-supplied text into realistic stroke trajectories. By leveraging lightweight 3D-printed materials and efficient mechanical designs, the system achieves a total hardware cost of approximately \56, significantly undercutting commercial alternatives. Experimental evaluations demonstrate handwriting precision within \pm$0.3 millimeters and a writing speed of approximately 200 mm/min, positioning the system as a viable solution for educational, research, and assistive applications. This study seeks to lower the barriers to personalized handwriting technologies, making them accessible to a broader audience.

The impressive performance of large language models (LLMs) on code-related tasks has shown the potential of fully automated software development. In light of this, we introduce a new software engineering task, namely Natural Language to code Repository (NL2Repo). This task aims to generate an entire code repository from its natural language requirements. To address this task, we propose a simple yet effective framework CodeS, which decomposes NL2Repo into multiple sub-tasks by a multi-layer sketch. Specifically, CodeS includes three modules: RepoSketcher, FileSketcher, and SketchFiller. RepoSketcher first generates a repository's directory structure for given requirements; FileSketcher then generates a file sketch for each file in the generated structure; SketchFiller finally fills in the details for each function in the generated file sketch. To rigorously assess CodeS on the NL2Repo task, we carry out evaluations through both automated benchmarking and manual feedback analysis. For benchmark-based evaluation, we craft a repository-oriented benchmark, SketchEval, and design an evaluation metric, SketchBLEU. For feedback-based evaluation, we develop a VSCode plugin for CodeS and engage 30 participants in conducting empirical studies. Extensive experiments prove the effectiveness and practicality of CodeS on the NL2Repo task.

In the creative practice of text-to-image (TTI) generation, images are synthesized from textual prompts. By design, TTI models always yield an output, even if the prompt contains unknown terms. In this case, the model may generate default images: images that closely resemble each other across many unrelated prompts. Studying default images is valuable for designing better solutions for prompt engineering and TTI generation. We present the first investigation into default images on Midjourney. We describe an initial study in which we manually created input prompts triggering default images, and several ablation studies. Building on these, we conduct a computational analysis of over 750,000 images, revealing consistent default images across unrelated prompts. We also conduct an online user study investigating how default images may affect user satisfaction. Our work lays the foundation for understanding default images in TTI generation, highlighting their practical relevance as well as challenges and future research directions.

A growing variety of prompt engineering techniques has been proposed for Large Language Models (LLMs), yet systematic evaluation of each technique on individual software engineering (SE) tasks remains underexplored. In this study, we present a systematic evaluation of 14 established prompt techniques across 10 SE tasks using four LLM models. As identified in the prior literature, the selected prompting techniques span six core dimensions (Zero-Shot, Few-Shot, Thought Generation, Ensembling, Self-Criticism, and Decomposition). They are evaluated on tasks such as code generation, bug fixing, and code-oriented question answering, to name a few. Our results show which prompting techniques are most effective for SE tasks requiring complex logic and intensive reasoning versus those that rely more on contextual understanding and example-driven scenarios. We also analyze correlations between the linguistic characteristics of prompts and the factors that contribute to the effectiveness of prompting techniques in enhancing performance on SE tasks. Additionally, we report the time and token consumption for each prompting technique when applied to a specific task and model, offering guidance for practitioners in selecting the optimal prompting technique for their use cases.

In this paper, we describe and present the first dataset of source code plagiarism specifically aimed at contest plagiarism. The dataset contains 251 pairs of plagiarized solutions of competitive programming tasks in Java, as well as 660 non-plagiarized ones, however, the described approach can be used to extend the dataset in the future. Importantly, each pair comes in two versions: (a) "raw" and (b) with participants' repeated template code removed, allowing for evaluating tools in different settings. We used the collected dataset to compare the available source code plagiarism detection tools, including state-of-the-art ones, specifically in their ability to detect contest plagiarism. Our results indicate that the tools show significantly worse performance on the contest plagiarism because of the template code and the presence of other misleadingly similar code. Of the tested tools, token-based ones demonstrated the best performance in both variants of the dataset.

We evaluate language models on their ability to explore interactive environments under a limited interaction budget. We introduce three parametric tasks with controllable exploration difficulty, spanning continuous and discrete environments. Across state-of-the-art models, we find systematic under-exploration and suboptimal solutions, with performance often significantly worse than simple explore--exploit heuristic baselines and scaling weakly as the budget increases. Finally, we study two lightweight interventions: splitting a fixed budget into parallel executions, which surprisingly improves performance despite a no-gain theoretical result for our tasks, and periodically summarizing the interaction history, which preserves key discoveries and further improves exploration.

This paper presents a study of using large language models (LLMs) in modifying existing code. While LLMs for generating code have been widely studied, their role in code modification remains less understood. Although "prompting" serves as the primary interface for developers to communicate intents to LLMs, constructing effective prompts for code modification introduces challenges different from generation. Prior work suggests that natural language summaries may help scaffold this process, yet such approaches have been validated primarily in narrow domains like SQL rewriting. This study investigates two prompting strategies for LLM-assisted code modification: Direct Instruction Prompting, where developers describe changes explicitly in free-form language, and Summary-Mediated Prompting, where changes are made by editing the generated summaries of the code. We conducted an exploratory study with 15 developers who completed modification tasks using both techniques across multiple scenarios. Our findings suggest that developers followed an iterative workflow: understanding the code, localizing the edit, and validating outputs through execution or semantic reasoning. Each prompting strategy presented trade-offs: direct instruction prompting was more flexible and easier to specify, while summary-mediated prompting supported comprehension, prompt scaffolding, and control. Developers' choice of strategy was shaped by task goals and context, including urgency, maintainability, learning intent, and code familiarity. These findings highlight the need for more usable prompt interactions, including adjustable summary granularity, reliable summary-code traceability, and consistency in generated summaries.

Large language models (LLMs) have revolutionized AI, but are constrained by limited context windows, hindering their utility in tasks like extended conversations and document analysis. To enable using context beyond limited context windows, we propose virtual context management, a technique drawing inspiration from hierarchical memory systems in traditional operating systems that provide the appearance of large memory resources through data movement between fast and slow memory. Using this technique, we introduce MemGPT (Memory-GPT), a system that intelligently manages different memory tiers in order to effectively provide extended context within the LLM's limited context window, and utilizes interrupts to manage control flow between itself and the user. We evaluate our OS-inspired design in two domains where the limited context windows of modern LLMs severely handicaps their performance: document analysis, where MemGPT is able to analyze large documents that far exceed the underlying LLM's context window, and multi-session chat, where MemGPT can create conversational agents that remember, reflect, and evolve dynamically through long-term interactions with their users. We release MemGPT code and data for our experiments at https://memgpt.ai.

Iterative RAG for multi-hop question answering faces challenges with lengthy contexts and the buildup of irrelevant information. This hinders a model's capacity to process and reason over retrieved content and limits performance. While recent methods focus on compressing retrieved information, they are either restricted to single-round RAG, require finetuning or lack scalability in iterative RAG. To address these challenges, we propose Notes Writing, a method that generates concise and relevant notes from retrieved documents at each step, thereby reducing noise and retaining only essential information. This indirectly increases the effective context length of Large Language Models (LLMs), enabling them to reason and plan more effectively while processing larger volumes of input text. Notes Writing is framework agnostic and can be integrated with different iterative RAG methods. We demonstrate its effectiveness with three iterative RAG methods, across two models and four evaluation datasets. Notes writing yields an average improvement of 15.6 percentage points overall, with minimal increase in output tokens.

Large language models (LLMs) have made significant progress in code generation tasks, but their performance in tackling programming problems with complex data structures and algorithms remains suboptimal. To address this issue, we propose an in-context learning approach that guides LLMs to debug by using a "print debugging" method, which involves inserting print statements to trace and analysing logs for fixing the bug. We collect a Leetcode problem dataset and evaluate our method using the Leetcode online judging system. Experiments with GPT-4 demonstrate the effectiveness of our approach, outperforming rubber duck debugging in easy and medium-level Leetcode problems by 1.5% and 17.9%.

When a human requests an LLM to complete a coding task using functionality from a large code repository, how do we provide context from the repo to the LLM? One approach is to add the entire repo to the LLM's context window. However, most tasks involve only fraction of symbols from a repo, longer contexts are detrimental to the LLM's reasoning abilities, and context windows are not unlimited. Alternatively, we could emulate the human ability to navigate a large repo, pick out the right functionality, and form a plan to solve the task. We propose MutaGReP (Mutation-guided Grounded Repository Plan Search), an approach to search for plans that decompose a user request into natural language steps grounded in the codebase. MutaGReP performs neural tree search in plan space, exploring by mutating plans and using a symbol retriever for grounding. On the challenging LongCodeArena benchmark, our plans use less than 5% of the 128K context window for GPT-4o but rival the coding performance of GPT-4o with a context window filled with the repo. Plans produced by MutaGReP allow Qwen 2.5 Coder 32B and 72B to match the performance of GPT-4o with full repo context and enable progress on the hardest LongCodeArena tasks. Project page: zaidkhan.me/MutaGReP

As Large Language Models (LLMs) are increasingly deployed as autonomous agents in complex and long horizon settings, it is critical to evaluate their ability to sabotage users by pursuing hidden objectives. We study the ability of frontier LLMs to evade monitoring and achieve harmful hidden goals while completing a wide array of realistic tasks. We evaluate a broad range of frontier LLMs using SHADE (Subtle Harmful Agent Detection & Evaluation)-Arena, the first highly diverse agent evaluation dataset for sabotage and monitoring capabilities of LLM agents. SHADE-Arena consists of complex pairs of benign main tasks and harmful side objectives in complicated environments. Agents are evaluated on their ability to complete the side task without appearing suspicious to an LLM monitor. When measuring agent ability to (a) complete the main task, (b) complete the side task, and (c) avoid detection, we find that the best performing frontier models score 27% (Claude 3.7 Sonnet) and 15% (Gemini 2.5 Pro) as sabotage agents when overseen by Claude 3.6 Sonnet. For current frontier models, success on the side task relies heavily on having access to a hidden scratchpad that is not visible to the monitor. We also use SHADE-Arena to measure models' monitoring abilities, with the top monitor (Gemini 2.5 Pro) achieving an AUC of 0.87 at distinguishing benign and malign transcripts. We find that for now, models still struggle at sabotage due to failures in long-context main task execution. However, our measurements already demonstrate the difficulty of monitoring for subtle sabotage attempts, which we expect to only increase in the face of more complex and longer-horizon tasks.

Large Language Models (LLMs) are rapidly transforming software engineering, with coding assistants embedded in an IDE becoming increasingly prevalent. While research has focused on improving the tools and understanding developer perceptions, a critical gap exists in understanding how developers actually use these tools in their daily workflows, and, crucially, where they struggle. This paper addresses part of this gap through a multi-phased investigation of developer interactions with an LLM-powered code editing and transformation feature, Transform Code, in an IDE widely used at Google. First, we analyze telemetry logs of the feature usage, revealing that frequent re-prompting can be an indicator of developer struggles with using Transform Code. Second, we conduct a qualitative analysis of unsatisfactory requests, identifying five key categories of information often missing from developer prompts. Finally, based on these findings, we propose and evaluate a tool, AutoPrompter, for automatically improving prompts by inferring missing information from the surrounding code context, leading to a 27% improvement in edit correctness on our test set.

Large Language Models (LLMs) demonstrate impressive capabilities across various domains, including role-playing, creative writing, mathematical reasoning, and coding. Despite these advancements, LLMs still encounter challenges with length control, frequently failing to adhere to specific length constraints due to their token-level operations and insufficient training on data with strict length limitations. We identify this issue as stemming from a lack of positional awareness and propose novel approaches--PositionID Prompting and PositionID Fine-Tuning--to address it. These methods enhance the model's ability to continuously monitor and manage text length during generation. Additionally, we introduce PositionID CP Prompting to enable LLMs to perform copy and paste operations accurately. Furthermore, we develop two benchmarks for evaluating length control and copy-paste abilities. Our experiments demonstrate that our methods significantly improve the model's adherence to length constraints and copy-paste accuracy without compromising response quality.

While voice user interfaces offer increased accessibility due to hands-free and eyes-free interactions, older adults often have challenges such as constructing structured requests and perceiving how such devices operate. Voice-first user interfaces have the potential to address these challenges by enabling multimodal interactions. Standalone voice + touchscreen Voice Assistants (VAs), such as Echo Show, are specific types of devices that adopt such interfaces and are gaining popularity. However, the affordances of the additional touchscreen for older adults are unknown. Through a 40-day real-world deployment with older adults living independently, we present a within-subjects study (N = 16; age M = 82.5, SD = 7.77, min. = 70, max. = 97) to understand how a built-in touchscreen might benefit older adults during device setup, conducting self-report diary survey, and general uses. We found that while participants appreciated the visual outputs, they still preferred to respond via speech instead of touch. We identified six design implications that can inform future innovations of senior-friendly VAs for managing healthcare and improving quality of life.

In today's software world with its cornucopia of reusable software libraries, when a programmer is faced with a programming task that they suspect can be completed through the use of a library, they often look for code examples using a search engine and then manually adapt found examples to their specific context of use. We put forward a vision based on a new breed of developer tools that have the potential to largely automate this process. The key idea is to adapt code autocompletion tools such that they take into account not only the developer's already-written code but also the intent of the task the developer is trying to achieve next, formulated in plain natural language. We call this practice of enriching the code with natural language intent to facilitate its completion natural language-guided programming. To show that this idea is feasible we design, implement and benchmark a tool that solves this problem in the context of a specific domain (data science) and a specific programming language (Python). Central to the tool is the use of language models trained on a large corpus of documented code. Our initial experiments confirm the feasibility of the idea but also make it clear that we have only scratched the surface of what may become possible in the future. We end the paper with a comprehensive research agenda to stimulate additional research in the budding area of natural language-guided programming.

User interface (UI) design is an iterative process in which designers progressively refine their work with design software such as Figma or Sketch. Recent advances in vision language models (VLMs) with tool invocation suggest these models can operate design software to edit a UI design through iteration. Understanding and enhancing this capacity is important, as it highlights VLMs' potential to collaborate with designers within conventional software. However, as no existing benchmark evaluates tool-based design performance, the capacity remains unknown. To address this, we introduce CANVAS, a benchmark for VLMs on tool-based user interface design. Our benchmark contains 598 tool-based design tasks paired with ground-truth references sampled from 3.3K mobile UI designs across 30 function-based categories (e.g., onboarding, messaging). In each task, a VLM updates the design step-by-step through context-based tool invocations (e.g., create a rectangle as a button background), linked to design software. Specifically, CANVAS incorporates two task types: (i) design replication evaluates the ability to reproduce a whole UI screen; (ii) design modification evaluates the ability to modify a specific part of an existing screen. Results suggest that leading models exhibit more strategic tool invocations, improving design quality. Furthermore, we identify common error patterns models exhibit, guiding future work in enhancing tool-based design capabilities.

Text-to-image generation models are powerful but difficult to use. Users craft specific prompts to get better images, though the images can be repetitive. This paper proposes a Prompt Expansion framework that helps users generate high-quality, diverse images with less effort. The Prompt Expansion model takes a text query as input and outputs a set of expanded text prompts that are optimized such that when passed to a text-to-image model, generates a wider variety of appealing images. We conduct a human evaluation study that shows that images generated through Prompt Expansion are more aesthetically pleasing and diverse than those generated by baseline methods. Overall, this paper presents a novel and effective approach to improving the text-to-image generation experience.

A number of tasks have been proposed recently to facilitate easy access to charts such as chart QA and summarization. The dominant paradigm to solve these tasks has been to fine-tune a pretrained model on the task data. However, this approach is not only expensive but also not generalizable to unseen tasks. On the other hand, large language models (LLMs) have shown impressive generalization capabilities to unseen tasks with zero- or few-shot prompting. However, their application to chart-related tasks is not trivial as these tasks typically involve considering not only the underlying data but also the visual features in the chart image. We propose PromptChart, a multimodal few-shot prompting framework with LLMs for chart-related applications. By analyzing the tasks carefully, we have come up with a set of prompting guidelines for each task to elicit the best few-shot performance from LLMs. We further propose a strategy to inject visual information into the prompts. Our experiments on three different chart-related information consumption tasks show that with properly designed prompts LLMs can excel on the benchmarks, achieving state-of-the-art.