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

Sep 30

Submitted by

ZhaoHaoyuu

MaLiang-Harness: A Programmable Path to Image and Video Generation

NationalUniversityofSingapore

National University of Singapore

Submitted by

rhfeiyang

Beyond the Timeline: Augmenting Long-Video Memory with Grounded Entity Biographies

amazon

Submitted by

LivXue

Raven: The Harness of Harnesses for Composable Agentic Intelligence

EverMindAI

Submitted by

rpzhou

What Makes World Action Models Generalize? An Empirical Study of Test-Time Future Modeling

Tsinghua-LeapLab

Tsinghua-LeapLab

Submitted by

vzl123

SAKI: Maximal-Coupling-Routed Teacher Supervision for On-Policy Distillation

meituan

Submitted by

jadohu

Learning Beyond What You Sample: Off-Policy-Aware Cross-Model Trajectory Exchange for RLVR

kaist-ai

Submitted by

DogNeverSleep

Think Before You Score: Thinking Reward Model for Visual Generation

·
12 authors

Submitted by

rkdrn79

Anisotropic Representations Improve Planning in JEPA World Models

SeoulNatlUniv

Seoul National University

Submitted by

AIcell

LEGO-Anything: Coding Agents for 3D Scene Reconstruction

AWS

Amazon Web Services

Submitted by

SteveZeyuZhang

WorldAttention: An Efficient Attention Architecture for Interactive Video World Models

Alibaba-DAMO-Academy

Submitted by

sanaka87

OmniTaskonomy: When Does Visual Generation Improve Visual Understanding?

Berkeley

Submitted by

desimfj

WorldLine: Action-Driven Visual Simulation for Robotic Manipulation

The Hong Kong University of Science and Technology

Submitted by

wchai

EasyPPO: Stabilizing the Critic Is Key

·
10 authors

Submitted by

Andyson

LongLive-Plug: Once-for-All Distillation for Video Generation

nvidia

Submitted by

Stephen-SMJ

When Does Dense Retrieval Need Asymmetric Geometry? A Bias-Variance Theory of Shared and Dual Projections

PolyUHK

The Hong Kong Polytechnic University

Submitted by

Ruiyang-061X

Marathoner: Ultra-Long-Horizon Autonomous Intelligence

antgroup

Submitted by

DogNeverSleep

HiRAE: Hierarchical Representation Autoencoding with Residual Budgets

·
8 authors

Submitted by

VINHYU

EVO-WAM: Evolving World Action Models through Video-Action Verification

hitsz123

Harbin Institute of Techonology, Shenzhen

Submitted by

Overdog

LIFT: Layout-In-Future Video Generation under Large Viewpoint Change via On-Policy Self-Distillation

UCSanDiego

University of California at San Diego

Submitted by

tuandattttt

CrossBFM: Distilling a Shared Latent Behavior Space Across Humanoid Embodiments

·
9 authors

Submitted by

cskrren

Real2Gym: Building Gyms from Videos, Bringing Skills to Robots

Shanghai-AI-Laboratory

Shanghai AI Laboratory

Submitted by

Jethro37

In-Context Learning for Robots: Methods and Applications

Submitted by

CFC

Where the Model Changes Its Mind: Hindsight-Divergence Localization for Efficient Reinforcement Learning with Verifiable Rewards

nvidia

Submitted by

lin1111987

TGRL: Temperature-Grouped Reinforcement Learning for Efficient Exploration in LLMs

meituan

Submitted by

LZXzju

HybridCUA: Learning to Orchestrate GUI and CLI for Computer-Use Agents

·
11 authors

Submitted by

Hiiamein

ROSS: Relearning from Self-Generated Rollouts through Selective Supervision

·
7 authors

Submitted by

LordNoah

AutoDataBench: Can Agents Write the Data That Feeds the Self-Improvement Loop?

·
8 authors

Submitted by

Jinfa

VideoLoop: Looped Working Memory Against Semantic Thrashing in Long-Form Video Agents

·
5 authors

Submitted by

zrrraa

Beyond Selection: Token Parameterization for Extreme Visual Token Compression

·
4 authors

Submitted by

xichenhku

PanoVLN: Towards Effective Panoramic Vision-and-Language Navigation

Zhejiang University

Submitted by

linxin02

Adversarial Training for Pixel Diffusion

adobe-research

Submitted by

FahdSeddik

Principled Thoughts for Latent Recursive LLM Systems

UBC-O

University of British Columbia - Okanagan Campus

Submitted by

0SilverBullet

Act First, Reason Later: Accelerating On-Policy Distillation for Multi-Turn Agents via Reference-Conditioned Inverse Dynamics

SouthernUniversityofScienceandTechnology

Southern University of Science and Technology

Submitted by

toshi2k2

FurE: Efficient Instance-Specific 3D Fur Reconstruction without Animal-Fur Datasets

·
4 authors

Submitted by

ForeverBlue

PrismQuant: Optimal Null-Space Rotations for Grouped Quantizers

NanyangTechnologicalUniversity

Nanyang Technological University