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

Feb 20

Submitted by

jt-zhang

SpargeAttention2: Trainable Sparse Attention via Hybrid Top-k+Top-p Masking and Distillation Fine-Tuning

Tsinghua University

5

Submitted by

xhyandwyy

Mobile-Agent-v3.5: Multi-platform Fundamental GUI Agents

AlibabaTongyiLab

Submitted by

taesiri

Unified Latents (UL): How to train your latents

google

Submitted by

jasonrqh

Frontier AI Risk Management Framework in Practice: A Risk Analysis Technical Report v1.5

AI45Research

4

Submitted by

taesiri

Arcee Trinity Large Technical Report

arcee-ai

Submitted by

johanneskirmayr

"What Are You Doing?": Effects of Intermediate Feedback from Agentic LLM In-Car Assistants During Multi-Step Processing

BMW-LLM-Research-Group

BMW LLM Research Group

Submitted by

wenwenD

Calibrate-Then-Act: Cost-Aware Exploration in LLM Agents

·
3 authors

Submitted by

taesiri

DDiT: Dynamic Patch Scheduling for Efficient Diffusion Transformers

amazon

Submitted by

youngw

TactAlign: Human-to-Robot Policy Transfer via Tactile Alignment

umich

University of Michigan

3

Submitted by

taesiri

Computer-Using World Model

·
18 authors

Submitted by

anuj0456

ArXiv-to-Model: A Practical Study of Scientific LM Training

KiteFishAI

Submitted by

taesiri

Discovering Multiagent Learning Algorithms with Large Language Models

google

Submitted by

JLiangHe

On the Mechanism and Dynamics of Modular Addition: Fourier Features, Lottery Ticket, and Grokking

Zhuoran Yang Research Group

Submitted by

han1997

FRAPPE: Infusing World Modeling into Generalist Policies via Multiple Future Representation Alignment

·
8 authors

Submitted by

gmongaras

2Mamba2Furious: Linear in Complexity, Competitive in Accuracy

smu

Southern Methodist University AI

Submitted by

pariard

CrispEdit: Low-Curvature Projections for Scalable Non-Destructive LLM Editing

UniversityofSouthernCalifornia

University of Southern California

Submitted by

yxzhang2024

World Models for Policy Refinement in StarCraft II

CASIA

Chinese Academic of Science Institute of Automation

Submitted by

oaishi

Modeling Distinct Human Interaction in Web Agents

Carnegie Mellon University School of Computer Science

Submitted by

JonasGeiping

NESSiE: The Necessary Safety Benchmark -- Identifying Errors that should not Exist

·
2 authors

Submitted by

daven3

Hardware Co-Design Scaling Laws via Roofline Modelling for On-Device LLMs

·
12 authors

Submitted by

SteveZeyuZhang

StereoAdapter-2: Globally Structure-Consistent Underwater Stereo Depth Estimation

PekingUniversity

Peking University

Submitted by

woorkhaarder

NeST: Neuron Selective Tuning for LLM Safety

is-tuda

Technical University of Darmstadt - Information Systems

Submitted by

henryL7

References Improve LLM Alignment in Non-Verifiable Domains

yale-nlp