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

Oct 15

Submitted by

lhmd

Any Time, Any Space: A World-State-Centric Framework for Agentic Video Generation

zju

Zhejiang University

Submitted by

kevin715

Beyond Domain-Routed Supervision: Cross-Teacher Guarding for Multi-Teacher On-Policy Distillation

jingdong1

1

Submitted by

yyyang

Player-1: Learning to Play from Synthetic Experience

TencentARC

ARC Lab, Tencent

Submitted by

YuyaoGe

Doc2Harness: Benchmarking LLMs on Long-Horizon Harness Engineering of Coding Agents

ict-cas

Institute of Computing Technology

Submitted by

kzhao5

Recovering Reliable Supervision via Critique-Guided On-Policy Distillation

·
7 authors

Submitted by

Ningyu

Cortex: Consolidating Context into Parametric Memory

ByteDance-Seed

1

Submitted by

yhx12

LatentLoop: Scaling Looped MoE Models with Asymmetric Compute Allocation

·
8 authors

Submitted by

zhongzero

More Methods, Better Agents? A Reality Check on Self-Distillation in Agentic Reinforcement Learning

zju

Zhejiang University

Submitted by

HuiningYuan

ActionOPD: Distilling Decisions, Not Tokens

nics-efc

Tsinghua-NICS-EFC

Submitted by

qingyuejiao

Seeing, Maintaining, and Learning: The Evolution of Multimodal Agent Memory

Fudan-University

Fudan University

Submitted by

yuUnuo

LensErase: A Foundation Model for Erase Anything

microsoft

Submitted by

sihwanpark

KV4-RL: Unlocking Faster Asynchronous Rollouts for RLVR with INT4 KV Cache Quantization

kaist-ai

Submitted by

KaituoFeng

Towards Agentic Robots: Model, Data, Environment, and Harness

·
16 authors

Submitted by

jiaruz2

Hybrid Reasoning Agents

StanfordUniversity

Stanford University

Submitted by

OfficerChul

HumanoidToolBench: Benchmarking Humanoid Tool Use from Selection to Mobile Execution

snupilab

Submitted by

Met4physics

Super-Resolution in The Right Latent Space: A Frozen Vision-Foundation Substrate

RTP-LLM

1

Submitted by

zsr200901

APT: Feedback-Driven Agentic Post-Training for Text-to-Image Generation

MicrosoftResearch

Microsoft Research

Submitted by

vxuanz

The Thinking Leakage in Reinforcement Learning for LLMs

kaust-generative-ai

KAUST Center of Excellence in Generative AI

Submitted by

xushuwen23

RuleTrek: Benchmarking Agentic Rule Reasoning in Complex Scenarios

·
13 authors

Submitted by

youganglyu

SwarmRSI: Recursive Self-Improvement through Evolving Scientific Agent Swarm

·
12 authors

Submitted by

sahsaeedi

τ-Emotion-Bench: Exposing the Interaction Fragility of LLM Agents through Real-World User Emotions

Arizona-State-University

Arizona State University

Submitted by

LazySheeep

RAR: Rate-Adaptive Rollout for Time-Consistent Interactive World Models

·
8 authors

Submitted by

AlanTang7

SkillMarker: A KGW-Based Watermarking Method for Agent Skill Ownership Verification

hitsz123

Harbin Institute of Techonology, Shenzhen

Submitted by

AIGCer-OPPO

GroundGUI: GUI Grounding by High-Quality Data and Distillation-to-Reinforcement Learning

·
8 authors

Submitted by

LLLeo612

ProgAgent: Learning Tool Orchestration through Programmatic Training

HKUST NLP Group

Submitted by

Flourish

Efficient Scene-Text Refinement via Pixel-Space Grouped Image Generation

ATH-MaaS

1

Submitted by

hannayukhymenko

Apertus 1.5: From Open Text Models to Open Multimodal Reasoners

swiss-ai

Swiss AI Initiative

1

Submitted by

HwanChang0106

Diagnosing and Correcting Reward Hacking in Verifiable Instruction Following

·
8 authors

Submitted by

linggm

Are We Testing Too Much? A Survey of Benchmark Compression for Efficient Evaluation of Large Language Models

SunYatsen

Sun Yat-Sen University

Submitted by

6cf

Long-Horizon RL Needs Muon

RUC

Renmin University of China

1

Submitted by

riccardodesanti

LLM-to-Flow Distillation: Teaching Natural Language Goals to Scientific Generators

ethz

Submitted by

ldkong

VITAL: From Visual Motion to Precise Action

HorizonRobotics

HorizonRobotics

Submitted by

ldkong

SafeDriveVLA: Navigation-Conditioned World Model Dreaming for Conflict-Aware End-to-End Autonomous Driving

UCI6055

University of California, Irvine

Submitted by

yantaiyang05

Acting Through Registers

Submitted by

zhuobai

Glance Forcing: Making Bidirectional Diffusion Autoregressive at a Glance

·
6 authors

Submitted by

jiwonsong

Joint Subspace Sparse Attention for Efficient Long-Context Decoding

SNU-VLSI

Seoul National University VLSI Lab

Submitted by

Flourish

TextVerifier: Seeing Beyond Recognition to Verify Chinese Text Rendering

ATH-MaaS

1

Submitted by

Ukpkmkkk

What Makes a Social Agent's Interactions Effective? Learning from Verifiable Outcomes in Paired Worlds

Tongyi-ConvAI

Submitted by

Lie24

StartLux-Decision: AutoThink Decision Models with Reinforcement Learning for Calibrated Deliberative Decisions

startlux-models

Submitted by

kostino

BigCell: Generating Gigapixel Whole-Slide Images

StonyBrook-CVLab

CVLab @ Stony Brook University

1

Submitted by

EVIGBYEN

SWEeper-Bench: Can Agents Discover Bugs in Interactive Software?

zlab-princeton

Submitted by

TechxGenus

Diff Diffusion: Predicting Edit Paths with Diffusion Models for Efficient Code Editing

·
8 authors

Submitted by

pengyue-polaron

Physical Time as a Missing Dimension of Robot Policy Memory Dynamics

newyorkuniversity

New York University

1

Submitted by

yantaiyang05

Exploring the Limits of Combinatorial Generalization in Embodied Manipulation

Submitted by

yflantmy

DecisionRobust: Benchmarking Robustness for JEV Style Decision Models Across Text and Vision

·
7 authors

1

Submitted by

akhadangi

Continuum Attention

·
1 authors

Submitted by

yumingfu

LabInstruct: Benchmarking Situated Instructional Video Generation for Lab Procedures

·
8 authors

1

Submitted by

Cii030

VASPO: Self-Distillation with Privileged Value Estimation

ustc-community

University of Science and Technology of China

Submitted by

FabianGroeger

To Align or Not To Align: Check Your COMPASS Before You Train

epfl-ch

Submitted by

FabianGroeger

Rethinking Projector Training in Multimodal LLMs

epfl-ch

Submitted by

rishii100

Rethinking LLM-Guided Visual Question Answering: Does Deeper Knowledge Integration Really Help?

·
5 authors

Submitted by

tksii

ProbabilityBench: Model Evaluation and Calibration with Exact Ground-Truth Uncertainty

ishidalab

Submitted by

kinpzz

Generative World Simulation with DAO and JING

XGEN-Labs

Submitted by

XINLI1997

Many Right Answers, One Training Example: Removing Responder Noise from GRPO for LLM Recommenders

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

XINLI1997

Share the Symbol, Not the Vector: Training-Free Fact Communication Across Language Models

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

Guan123

SyncReward: Human-Aligned Synchronization Reward for Audio-Visual Generation

tencent

Submitted by

PeterLauLukCh

Scaling Online Optimal Transport Minibatches for Flow Matching

UCBerkeley

University of California, Berkeley

1

Submitted by

chen-yingfa

Latent Compaction with Test-Time Training

TsinghuaNLP

Tsinghua NLP Group

Submitted by

ZGZzz

NavGPT-3: Harnessing Context in a Hierarchical Navigation Runtime

Metacognition-AI

Submitted by

Shuaijun

MimicX: Policy-in-the-Loop Supervision Refinement for Video-Driven Humanoid Motion Tracking

·
10 authors

Submitted by

IanMagnusson

The COUNT: Counting N-grams Unlocks Scaling on Host Memory and Training-Free Adaptation

University of Washington

Submitted by

CliverChen

Training LLMs for Multi-Agent Collaboration with Verifiable Information Rewards

·
5 authors

1

Submitted by

walterdev

Your Transformer Is Secretly Elastic

VISLab-Amsterdam

Video & Image Sense Lab

Submitted by

zhuobai

StreamErase: Towards Ultra-Fast Streaming Video Erasing

·
7 authors

Submitted by

lalalamdbf

AgentSafeEvo: Adversarial Co-Evolution for LLM Agent Safety

·
7 authors

1

Submitted by

Tangerine24

Vela: Scaling Vision-Language-Action Models with Adaptive Action Curve Parametrization

·
7 authors

Submitted by

xyzzzh

Jag: Direct Box Prediction for Efficient Visual Grounding

·
6 authors

Submitted by

pengyue-polaron

G^3VLA: Geometric Inductive Bias for Vision-Language-Action Models

·
8 authors

1

Submitted by

taesiri

MA-Bench: Measuring Operational Planning in Language Models

·
4 authors