new

Get trending papers in your email inbox once a day!

Get trending papers in your email inbox!

Daily Papers

byAK and the research community

Jun 1

Submitted by

alireza7

GrepSeek: Training Search Agents for Direct Corpus Interaction

University of Massachusetts Amherst

Submitted by

jasonrqh

COLLEAGUE.SKILL: Automated AI Skill Generation via Expert Knowledge Distillation

Submitted by

Myashka

Trust-Region Behavior Blending for On-Policy Distillation

t-tech

1

Submitted by

taesiri

Representation Forcing for Bottleneck-Free Unified Multimodal Models

ByteDance-Seed

Submitted by

AaronZ345

SwanVoice: Expressive Long-Form Zero-Shot Speech Synthesis for Both Monologue and Dialogue

ByteDance

1

Submitted by

NeoZ123

LongTraceRL: Learning Long-Context Reasoning from Search Agent Trajectories with Rubric Rewards

THU-KEG

Knowledge Engineer Group @ Tsinghua University

Submitted by

Ar4l

Mellum2 Technical Report

JetBrains

Submitted by

VCLab-PolyU

GGT-100K: Generative Ground Truth for Generalizable Real-World Image Restoration

VCLab-HKPU

Submitted by

yanxg

Function2Scene: 3D Indoor Scene Layout from Functional Specifications

·
7 authors

Submitted by

AaronZ345

Towards Streaming Synchronized Spatial Audio Generation via Autoregressive Diffusion Transformer

zju

Zhejiang University

1

Submitted by

Yuyang-z

SANA-Streaming: Real-time Streaming Video Editing with Hybrid Diffusion Transformer

nvidia

Submitted by

AaronZ345

Comprehensive Benchmarking of Long-Form Speech Generation in Diverse Scenarios

zju

Zhejiang University

1

Submitted by

hyc2026

Task-Focused Memorization for Multimodal Agents

ByteDance-Seed

Submitted by

FSCCS

dMoE: dLLMs with Learnable Block Experts

·
5 authors

Submitted by

wyy-code

Not All Disagreement Is Learnable: Token Teachability in On-Policy Distillation

PolyUHK

The Hong Kong Polytechnic University

Submitted by

231sm

Exploring Autonomous Agentic Data Engineering for Model Specialization

Zhejiang University

Submitted by

nelikCode

PEEK: Picking Essential frames via Efficient Knowledge distillation

momentslab

Submitted by

pminervini

SCOPE: Self-Play via Co-Evolving Policies for Open-Ended Tasks

·
4 authors

1

Submitted by

liussy

SAAS: Self-Aware Reinforcement Learning for Over-Search Mitigation in Agentic Search

XMU

Xiamen University

Submitted by

smallnono

Recovering Policy-Induced Errors: Benchmarking and Trajectory Synthesis for Robust GUI Agents

·
10 authors

Submitted by

zstanjj

From Prompt Injection to Persistent Control: Defending Agentic Harness Against Trojan Backdoors

·
7 authors

Submitted by

Ningyu

LongDS-Bench: On the Failure of Long-Horizon Agentic Data Analysis

antgroup

Submitted by

zcai

VLM3: Vision Language Models Are Native 3D Learners

Submitted by

JeffreyYzh

DecMem: Towards Minute-Long Consistent World Generation with Decoupled Memory

KlingTeam

Submitted by

taesiri

Linear Scaling Video VLMs for Long Video Understanding

StanfordUniversity

Stanford University

Submitted by

sam121796

Hide-and-Seek in Trajectories: Discovering Failure Signals for VLA Runtime Monitoring

Uwmadison

University of Wisconsin-Madison

Submitted by

Gigglingface

How can embedding models bind concepts?

·
3 authors

Submitted by

jhying

OpenSkillEval: Automatically Auditing the Open Skill Ecosystem for LLM Agents

·
5 authors

Submitted by

EvilScript

Emergent Languages in Populations of Language Model Agents: From Token Efficiency to Oversight Evasion

aisilab

AI Safety & Interpretability Lab

Submitted by

Yofuria

The Flip Side of RLHF: On-Policy Feedback for Reward Model Self-Supervised Improvement

·
6 authors

1

Submitted by

jky594176

GDSD: Reinforcement Learning as Guided Denoiser Self-Distillation for Diffusion Language Models

·
7 authors

1

Submitted by

Yuezhangjoslin

Seeing Isn't Knowing: Do VLMs Know When Not to Answer Spatial Questions (and Why)?

·
6 authors

Submitted by

danjacobellis

FRAPPE: Full Input, Residual Output Autoencoding with Projection Pursuit Encoder

·
2 authors

Submitted by

justachetan

Flat-Pack Bench: Evaluating Spatio-Temporal Understanding in Large Vision-Language Models through Furniture Assembly

·
8 authors

Submitted by

mujianijan

DRIFT: Decoupled Rollouts and Importance-Weighted Fine-Tuning for Efficient Multi-Turn Optimization

·
5 authors

Submitted by

taesiri

Lumos-Nexus: Efficient Frequency Bridging with Homogeneous Latent Space for Video Unified Models

·
12 authors

Submitted by

MengqiLei

Count Anything

·
7 authors

Submitted by

hosytuyen

SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?

Furong Huang's Lab at UMD

Submitted by

amanchadha

MAAT: Multi-phase Adapter-Aware Targeted Unlearning

·
6 authors

1

Submitted by

fengnian1678

VisualThink-VLA: Visual Intermediate Reasoning for Effective and Low-Latency Vision-Language-Action Policies

ZhejiangDun

Zhejiang university

Submitted by

Shuwan

The Good, the Bad, and the Ugly of Markov Boundary for Tabular Prediction

CSE472-blanket-challenge

Beyond the Blanket

1

Submitted by

JwonP

When Confidence Misleads: Suffix Anchoring and Anchor-Proximity Confidence Modulation for Diffusion Language Models

·
5 authors

Submitted by

koyy001

HL-OutPaint: Coarse-to-Fine Video Outpainting for High-Resolution Long-Range Videos

postech-cglab

POSTECH Computer Graphics Lab

1

Submitted by

allencbzhang

iVGR: Internalizing Visually Grounded Reasoning for MLLMs with Reinforcement Learning

HKUCDS

University of Hong Kong

Submitted by

taesiri

Light Interaction: Training-Free Inference Acceleration for Interactive Video World Models

nvidia

Submitted by

shjo-april

One Click per Cell Type Suffices: Training-free Group Interaction for Cell Instance Segmentation

SeoulNatlUniv

Seoul National University

Submitted by

Shangding-Gu

From Model Scaling to System Scaling: Scaling the Harness in Agentic AI

Berkeley

Submitted by

billpsomas

Benchmarking Composed Image Retrieval for Applied Earth Observation

vrg-prague

Visual Recognition Group FEE CTU in Prague

Submitted by

stefan-baumann

RayDer: Scalable Self-Supervised Novel View Synthesis from Real-World Video

CompVis

Submitted by

josefchen

Memory-Bound but Not Bandwidth-Limited: The Physical AI Inference Gap in Batch-1 LLM Decode

·
1 authors

1

Submitted by

L-yiheng

AnyMo: Scaling Any-Modality Conditional Motion Generation with Masked Modeling

·
7 authors

Submitted by

HenryWJL

Frequency-Guided Action Diffusion via Sub-Frequency Manifold Traversal

upenn

University of Pennsylvania

Submitted by

Braudo

Beyond Holistic Models: Systematic Component-level Benchmarking of Deep Multivariate Time-Series Forecasting

·
7 authors

Submitted by

jye9

A Topology-Aware Spatiotemporal Handover Framework for Continuous Multi-UAV Tracking

·
3 authors