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

LivXue

Raven: The Harness of Harnesses for Composable Agentic Intelligence

EverMindAI

Submitted by

ZhaoHaoyuu

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

NationalUniversityofSingapore

National University of Singapore

Submitted by

Jethro37

In-Context Learning for Robots: Methods and Applications

Submitted by

yanlinli

Omni-IO Skills: Harnessing Your Agent Omni-Native

NationalUniversityofSingapore

National University of Singapore

Submitted by

AustinXiao

VoxMem: Benchmarking Multimodal Memory in Large Audio Language Models

unimelb-hf

The University of Melbourne

Submitted by

xichenhku

PanoVLN: Towards Effective Panoramic Vision-and-Language Navigation

Zhejiang University

1

Submitted by

AIcell

LEGO-Anything: Coding Agents for 3D Scene Reconstruction

AWS

Amazon Web Services

1

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

DogNeverSleep

Think Before You Score: Thinking Reward Model for Visual Generation

·
12 authors

Submitted by

rhfeiyang

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

AmazonScience

Submitted by

starsuzi

Follow the Entities: A Corpus Map for Agentic Search

microsoft

Submitted by

MrDarkTesla

LLMs are General Asynchronous Agents

yresearch

Yandex Research

Submitted by

torpedo2003

EngiWorld: What Can Frontier Agents Deliver in Professional Engineering Environments?

zhiman-ai

Submitted by

zihan-audio

Beyond Dyadic Memory: Interaction-Aware Multimodal Memory with Adaptive Agentic Retrieval for Multi-Party Spoken Conversations

Zhejiang University

Submitted by

sanaka87

OmniTaskonomy: When Does Visual Generation Improve Visual Understanding?

Berkeley

Submitted by

taesiri

LongCat-DeepResearch Technical Report

meituan-longcat

Submitted by

jadohu

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

kaist-ai

1

Submitted by

rkdrn79

Anisotropic Representations Improve Planning in JEPA World Models

SeoulNatlUniv

Seoul National University

Submitted by

Owen777

SoL-Refiner: Speed-of-Light One-Step Refinement for High-Resolution Video

·
12 authors

Submitted by

Jianguo-Huang11

APM-Bench: Benchmarking Cross-session Persistent Memory for Egocentric Streaming Video Assistants

·
12 authors

Submitted by

dolev31

Asking for What Was Never Requested: Horizontal and Vertical Proactivity in Agents

ibm

Submitted by

Andyson

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

nvidia

Submitted by

LZXzju

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

·
11 authors

Submitted by

VINHYU

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

hitsz123

Harbin Institute of Techonology, Shenzhen

Submitted by

Hiiamein

ROSS: Relearning from Self-Generated Rollouts through Selective Supervision

AllSpark-Research

AllSpark Research

1

Submitted by

tuandattttt

CrossBFM: Distilling a Shared Latent Behavior Space Across Humanoid Embodiments

·
9 authors

1

Submitted by

SteveZeyuZhang

WorldAttention: An Efficient Attention Architecture for Interactive Video World Models

Alibaba-DAMO-Academy

Submitted by

wchai

EasyPPO: Stabilizing the Critic Is Key

·
10 authors

Submitted by

Ruiyang-061X

Marathoner: Ultra-Long-Horizon Autonomous Intelligence

antgroup

Submitted by

rulins

Context Language Models

meta

Submitted by

desimfj

WorldLine: Action-Driven Visual Simulation for Robotic Manipulation

The Hong Kong University of Science and Technology

Submitted by

DogNeverSleep

HiRAE: Hierarchical Representation Autoencoding with Residual Budgets

·
8 authors

1

Submitted by

zfz04

FocusVTC: Efficient and High-Performance Visual Text Compression with Adaptive Resolution

·
7 authors

Submitted by

nishadsinghi

Reasoning with Image Generation

MAI-Lab

Multimodal AI Lab - Technical University of Darmstadt

Submitted by

teddybearagee

ANTMAN: Adaptive Need Tracking for Multi-Agent Navigation in Large Information Spaces

UIUC-CS

University of Illinois at Urbana-Champaign

1

Submitted by

shim0114

AutoRef: Harness Optimization for Agentic Multi-Reference Image Generation

·
6 authors

Submitted by

EvilScript

Selecting The Most Informative Tokens in Natural Language Autoencoders

aisilab

AI Safety & Interpretability Lab

Submitted by

ElvisWang111

Scheduling Recursive Reasoning in Looped Transformers

·
6 authors

1

Submitted by

MBJinX

Omni-Decision: Evidence-Ledger Planning for Omni-Modal Agents

·
9 authors

Submitted by

deqing

TabFM: A Zero-Shot Foundation Model for Tabular Data

google

1

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

Gene-Liu

Routing Should Pay for Itself: Sparse Supervision for Economical LLM Routing

SinapisAI

Submitted by

kph68

EmoRES-TTS: Residual-Enhanced Vector Steering for Emotional Speech Generation

meta

Submitted by

ZhaoHaoyuu

Chinese-Jev: Bringing System One Model to Chinese-Language Tasks

Fudan-University

Fudan University

Submitted by

Jinfa

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

·
5 authors

Submitted by

linxin02

Adversarial Training for Pixel Diffusion

adobe-research

Submitted by

deqing

TabFM-Auto: Self-Evolving Pipelines for Tabular Foundation Models

google

1

Submitted by

lin1111987

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

meituan

Submitted by

yeyingjin

StoryEngine: A State-Grounded Agentic Framework for Video Storytelling

NationalUniversityofSingapore

National University of Singapore

Submitted by

zengziyun

EpiCon: Collective Agent Learning through Co-Evolving Multimodal Memory

·
6 authors

1

Submitted by

taesiri

Language Models Are "Insecure" Reporters

google

Submitted by

zrrraa

Beyond Selection: Token Parameterization for Extreme Visual Token Compression

·
4 authors

Submitted by

cskrren

Real2Gym: Building Gyms from Videos, Bringing Skills to Robots

Shanghai-AI-Laboratory

Shanghai AI Laboratory

Submitted by

CFC

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

nvidia

Submitted by

RobinRoaR

PMOPD: Task Ordering, Cycling, and Parameter-Update Subspace Protection in Multi-Teacher On-Policy Distillation

antgroup

Submitted by

tommymarto

Improved Distributional Diffusion Models

CompVis

Submitted by

ruihugface

AnyStep-WAM: Budget-Aligned Distillation and Adaptive Inference for World Action Models

·
7 authors

Submitted by

YUXU915

Breaking the Uniformity Trap: Scaling Video Diffusion Model via SplitMoE

ByteDance

1

Submitted by

shayekh

PlaylistEval: Can Video-Language Judges Be Trusted at Day Scale and Beyond?

·
2 authors

1

Submitted by

LordNoah

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

·
8 authors

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

Huxaifa

WISE-ATTA: When to Ask for Labels in Budgeted Active Test-Time Adaptation

cispa

CISPA – Helmholtz-Zentrum für Informationssicherheit gGmbH

Submitted by

ChongWang1024

Persistence Forcing: Exploiting Feature Specialization in Pixel-Space Diffusion

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

Frinkleko

SEAD: A State-Based Perspective on Attack and Defense in Tool-Using Agents

·
4 authors

Submitted by

sangminwoo

StructRL: Online Structured Reinforcement Learning for Long-Horizon Vision-Language-Action Tasks

amazon

Submitted by

YanZhanPKU

Same Bytes, Different Authority: Reserved-Token Representations in Chat-Template Prompt Injection

PekingUniversity

Peking University

Submitted by

ArgonnSZ

Pretraining Transformers with Quantized Softmax in Attention

illinois

University of Illinois at Urbana-Champaign

1

Submitted by

crs25-tsinghua

One Proposal for Every Margin: Zero-Shot Amortized Sequential Importance Sampling for Binary Matrices

Tsinghua University

1

Submitted by

blue531

Preference-Guided Adaptation for Open-Vocabulary Semantic Segmentation via Prompt Disagreement

Submitted by

ForeverBlue

PrismQuant: Optimal Null-Space Rotations for Grouped Quantizers

NanyangTechnologicalUniversity

Nanyang Technological University

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

1

Submitted by

toshi2k2

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

·
4 authors

Submitted by

dwenlong

Learning Dynamics of Continual Learning: A Unified View of Data Attribution, Forgetting, and Plasticity Loss

·
5 authors

Submitted by

GiulioFranzese

ALICE: In-context, Zero-shot, Mutual Information Estimation

eurecom-probai

Probabilistic AI and Learning Group @EURECOM

Submitted by

thomas-ferraz

Scaffolding Minds: Optimizing Latent Visual Target Representations for Multimodal Reasoning

GoogleDeepMind

Submitted by

RPorcedda

Jev thinks "I don't know'', but doesn't say it: Introducing Sys1Cal-v1 Dataset for Probability Calibration

·
1 authors

Submitted by

zichao1

Understanding On-Policy Distillation: A Mechanistic Interpretability Perspective via Sparse Crosscoders

hkuhk

The University of Hong Kong

Submitted by

colored-dye

Scaling Properties of Same-Family On-Policy Distillation

zju

Zhejiang University

2

Submitted by

sukriramli

Hyperspherical Semantic Trajectory Analysis: Mapping Technological Diffusion across Academic Preprints, Patent Signals, and Compute Scaling

·
1 authors

Submitted by

ZHEN-04

CaptchaArena: A Large-Scale, Fine-Grained Dataset for Training Computer-Use Agents on Interactive CAPTCHAs

ColumbiaUniversity

Columbia University