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

May 24-30

Submitted by

taesiri

Gamma-World: Generative Multi-Agent World Modeling Beyond Two Players

nvidia

Submitted by

taesiri

SkillOpt: Executive Strategy for Self-Evolving Agent Skills

MicrosoftResearch

Microsoft Research

Submitted by

Nothing2Say

DVAO: Dynamic Variance-adaptive Advantage Optimization for Multi-reward Reinforcement Learning

·
6 authors

3

Submitted by

taesiri

LocateAnything: Fast and High-Quality Vision-Language Grounding with Parallel Box Decoding

nvidia

4

Submitted by

Met4physics

Rethinking Cross-Layer Information Routing in Diffusion Transformers

RTP-LLM

5

Submitted by

Jinjing713

Lens: Rethinking Training Efficiency for Foundational Text-to-Image Models

microsoft

Submitted by

Kaining

WBench: A Comprehensive Multi-turn Benchmark for Interactive Video World Model Evaluation

meituan-longcat

Submitted by

shenqiorient

AgentDoG 1.5: A Lightweight and Scalable Alignment Framework for AI Agent Safety and Security

Submitted by

Mithas-01

ProRL: Effective Reinforcement Learning for Proactive Recommendation via Rectified Policy Gradient Estimation

Fudan-University

Fudan University

Submitted by

anchen1011

Macaron-A2UI: A Model for Generative UI in Personal Agents

mindlab-research

5

Submitted by

taesiri

Agent Explorative Policy Optimization for Multimodal Agentic Reasoning

nvidia

4

Submitted by

EddieYang428

EvalVerse: Pipeline-Aware and Expert-Calibrated Benchmarking for Professional Cinematic Video Generation

tencent

Submitted by

Bang-UdeM-Mila

Foundation Protocol: A Coordination Layer for Agentic Society

University of Montreal

Submitted by

taesiri

Qwen-VLA: Unifying Vision-Language-Action Modeling across Tasks, Environments, and Robot Embodiments

Qwen

Submitted by

Paranioar

From Pixels to Words -- Towards Native One-Vision Models at Scale

·
21 authors

3

Submitted by

lifuguan

SpatialBench: Is Your Spatial Foundation Model an All-Round Player?

ropedia-ai

Submitted by

Abyssaledge

MobileGym: A Verifiable and Highly Parallel Simulation Platform for Mobile GUI Agent Research

·
11 authors

Submitted by

Ningyu

SciAtlas: A Large-Scale Knowledge Graph for Automated Scientific Research

University College London CS

Submitted by

jinheon

OmniRetrieval: Unified Retrieval across Heterogeneous Knowledge Sources

kaist-ai

Submitted by

Xkev

Self-Improving Language Models with Bidirectional Evolutionary Search

Harvard

Harvard University

Submitted by

lhmd

TriSplat: Simulation-Ready Feed-Forward 3D Scene Reconstruction

zju

Zhejiang University

Submitted by

jamesliu1217

CollectionLoRA: Collecting 50 Effects in 1 LoRA via Multi-Teacher On-Policy Distillation

·
10 authors

Submitted by

giantPanda0906

StepAudio 2.5 Technical Report

·
101 authors

Submitted by

SII-Molu

DenoiseRL: Bootstrapping Reasoning Models to Recover from Noisy Prefixes

Fudan-University

Fudan University

Submitted by

amphora

ResearchMath-14K: Scaling Research-Level Mathematics via Agents

SeoulNatlUniv

Seoul National University

2

Submitted by

HansonDJN

Toward Native Multimodal Modeling: A Roadmap

tencent

Submitted by

taesiri

PiD: Fast and High-Resolution Latent Decoding with Pixel Diffusion

nvidia

Submitted by

taesiri

minWM: A Full-Stack Open-Source Framework for Real-Time Interactive Video World Models

·
12 authors

Submitted by

joesharratt29

ThriftAttention: Selective Mixed Precision for Long-Context FP4 Attention

·
1 authors

Submitted by

jinlovespho

Geometry-Aware Representation Denoising for Robust Multi-view 3D Reconstruction

kaist-ai

Submitted by

Yeonjun

Personalize-then-Store: Benchmarking and Learning Personalized Memory for Long-horizon Agents

Submitted by

Zuyan

GEM: Generative Supervision Helps Embodied Intelligence

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

Ningyu

MemTrace: Tracing and Attributing Errors in Large Language Model Memory Systems

alibaba-inc

Submitted by

DogNeverSleep

LongAV-Compass: Towards Unified Evaluation of Minute-Scale Audio-Visual Generation Across T2AV, I2AV, and V2AV

KlingTeam

Submitted by

hsaest

QUEST: Training Frontier Deep Research Agents with Fully Synthetic Tasks

osunlp

Submitted by

KangsanKim71

Learn from Weaknesses: Automated Domain Specialization for Small Computer-Use Agents

kaist-ai

Submitted by

Samchen374

D^2-Monitor: Dynamic Safety Monitoring for Diffusion LLMs via Hesitation-Aware Routing

Oxford

University of Oxford

2

Submitted by

mwxely

ParaVT: Taming the Tool Prior Paradox for Parallel Tool Use in Agentic Video Reinforcement Learning

lmms-lab

Submitted by

yulunliu

YoCausal: How Far is Video Generation from World Model? A Causality Perspective

ShandaAI

Submitted by

BBBBCHAN

See What I Mean: Aligning Vision and Language Representations for Video Fine-grained Object Understanding

AlibabaTongyiLab

Submitted by

taesiri

The MiniMax-M2 Series: Mini Activations Unleashing Max Real-World Intelligence

MiniMaxAI

Submitted by

memray

ScientistOne: Towards Human-Level Autonomous Research via Chain-of-Evidence

google

Submitted by

KevinQHLin

Soap2Soap: Long Cinematic Video Remaking via Multi-Agent Collaboration

showlab

Submitted by

TheMartyr

JLT: Clean-Latent Prediction in Latent Diffusion Transformers

dawn-neo

Submitted by

BryanWangNLP

CUA-Gym: Scaling Verifiable Training Environments and Tasks for Computer-Use Agents

Qwen

Submitted by

tgy2024

AutoResearch AI: Towards AI-Powered Research Automation for Scientific Discovery

·
23 authors

Submitted by

taesiri

From Raw Experience to Skill Consumption: A Systematic Study of Model-Generated Agent Skills

MicrosoftResearch

Microsoft Research

Submitted by

bitwxl2022

Share More, Search Less: Collaborative Parallel Thinking for Efficient Test-Time Scaling

BIT

Beijing Institute of Technology

Submitted by

Zfancy

Your Agents Are Aging Too: Agent Lifespan Engineering for Deployed Systems

·
8 authors

Submitted by

SereinH

GenClaw: Code-Driven Agentic Image Generation

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

Zuica96

PhotoFlow: Agentic 3D Virtual Photography Missions

Visionary-Laboratoary

Visionary-Laboratoary

Submitted by

xiangan

LLaVA-OneVision-2: Towards Next-Generation Perceptual Intelligence

·
30 authors

Submitted by

HanSolo9682

Your Embedding Model is SMARTer Than You Think

·
6 authors

Submitted by

Ningyu

Rethinking Memory as Continuously Evolving Connectivity

alibaba-inc

2

Submitted by

yflantmy

SkillGrad: Optimizing Agent Skills Like Gradient Descent

·
5 authors

Submitted by

zino1

VGenST-Bench: A Benchmark for Spatio-Temporal Reasoning via Active Video Synthesis

·
4 authors

Submitted by

Haiyang-W

Claw-Anything: Benchmarking Always-On Personal Assistants with Broader Access to User's Digital World

·
11 authors

Submitted by

yixuantt

AI Research Agents Narrow Scientific Exploration

·
2 authors

Submitted by

Ningyu

How LoRA Remembers? A Parametric Memory Law for LLM Finetuning

alibaba-inc

Submitted by

jonathanShi

UniSteer: Text-Guided Flow Matching in Activation Space for Versatile LLM Steering

ShanghaiTech

ShanghaiTech University

Submitted by

yunyangge

OSP-Next: Efficient High-Quality Video Generation with Sparse Sequence Parallelism, HiF8 Quantization, and Reinforcement Learning

PekingUniversity

Peking University

Submitted by

Koi953215

Pantheon360: Taming Digital Twin Generation via 3D-Aware 360° Video Diffusion

BoschUS

Submitted by

robingg1

Native Audio-Visual Alignment for Generation

baidu

Submitted by

taesiri

MUSE-Autoskill: Self-Evolving Agents via Skill Creation, Memory, Management, and Evaluation

ByteDance

Submitted by

jhhuangchloe

Recursive Flow Matching

·
4 authors

2

Submitted by

dericky286

ControlLight: Towards Controllable, Consistent, and Generalizable Low-Light Enhancement

·
7 authors

Submitted by

Parkprogrammer

Self-Improving CAD Generation Agents with Finite Element Analysis as Feedback

SeoulNatlUniv

Seoul National University

2

Submitted by

rookiexiong

LoMo: Local Modality Substitution for Deeper Vision-Language Fusion

Fudan-University

Fudan University

Submitted by

bowiehsu

Long Live The Balance: Information Bottleneck Driven Tree-based Policy Optimization

alibaba-inc

Submitted by

nielsr

Gemini Embedding 2: A Native Multimodal Embedding Model from Gemini

google

Submitted by

yang29

On-Policy Adversarial Flow Distillation for Autoregressive Video Generation

·
7 authors

3

Submitted by

donghao-zhou

SkillEvolBench: Benchmarking the Evolution from Episodic Experience to Procedural Skills

ohiostate

The Ohio State University

Submitted by

Concyclics

MemForest: An Efficient Agent Memory System with Hierarchical Temporal Indexing

·
9 authors

Submitted by

talzoomanzoo

LaRA: Layer-wise Representation Analysis for Detecting Data Contamination in RL Post-Training

yonseiworld

Yonsei University

1

Submitted by

chanhee-luke

Why Far Looks Up: Probing Spatial Representation in Vision-Language Models

nvidia

Submitted by

heroding77

Skill0.5: Joint Skill Internalization and Utilization for Out-of-Distribution Generalization in Agentic Reinforcement Learning

ECNU

East China Normal University

Submitted by

QianYangMILA

How and What to Imagine? Visual Thinking in Unified Multimodal Models for Cross-View Spatial Reasoning

Submitted by

taesiri

Negligible in Size, Significant in Effect: On Scale Vectors in Large Language Models

ByteDance-Seed

Submitted by

xiwenyoumu

Fast-dDrive: Efficient Block-Diffusion VLM for Autonomous Driving

nvidia

Submitted by

chengq9

Advancing Creative Physical Intelligence in Large Multimodal Models

UIUC-CS

University of Illinois at Urbana-Champaign

Submitted by

Alex7616

Anticipate and Learn: Unleashing Idle-Time Compute in Proactive Agents

SJTU

Shanghai Jiao Tong University

Submitted by

CircleRadon

InstructSAM: Segment Any Instance with Any Instructions

·
9 authors

Submitted by

INV-WZQ

SCOPE: Simulating Cross-game Operations in Playable Environments for FPS World Models

·
14 authors

Submitted by

Ningyu

When Should Models Change Their Minds? Contextual Belief Management in Large Language Models

Zhejiang University

Submitted by

ryancll118

GE-Sim 2.0: A Roadmap Towards Comprehensive Closed-loop Video World Simulators for Robotic Manipulation

agibot-world

Submitted by

Chen1999

VitaBench 2.0: Evaluating Personalized and Proactive Agents in Long-Term User Interactions

meituan-longcat

Submitted by

lilvjosephtang

RankJudge: A Multi-Turn LLM-as-a-Judge Synthetic Benchmark Generator

Layer6

Submitted by

hyeoncho01

Balancing Fidelity and Diversity in Diffusion Models via Symmetric Attention Decomposition: Hopfield Perspective

·
3 authors

Submitted by

CuSO4-Chen

Efficient Agentic Reinforcement Learning with On-Policy Intrinsic Knowledge Boundary Enhancement

tencent

Submitted by

domejiraphon

Helix4D: Complex 4D Mesh Generation

snap-research

3

Submitted by

Songweii

Channel-wise Vector Quantization

·
8 authors

Submitted by

yoavgur

Faithfulness Metrics Don't Measure Faithfulness: A Meta-Evaluation with Ground Truth

tau

Tel Aviv University

Submitted by

yasNing

HRBench: Benchmarking and Understanding Thinking-Mode Switch Strategies in Hybrid-Reasoning LLMs

tencent

Submitted by

Afeng-x

Rethinking VLM Representation for VLA Initialization

CUHK-CSE

The Chinese University of Hong Kong

Submitted by

yuhangzang

ETCHR: Editing To Clarify and Harness Reasoning

internlm

Intern Large Models

Submitted by

NoamIssachar

Colored Noise Diffusion Sampling

The Hebrew University of Jerusalem

Submitted by

taesiri

CubePart: An Open-Vocabulary Part-Controllable 3D Generator

·
12 authors

Submitted by

seungyoonee

Is Position Bias in Dense Retrievers Built In-or Learned from Data?

sionic-ai

Submitted by

Chen1999

Learning to Act under Noise: Enhancing Agent Robustness via Noisy Environments

meituan-longcat

Submitted by

Jiasenlu

MobileMoE: Scaling On-Device Mixture of Experts

Submitted by

AuroraZengfh

Does Seeing More Mean Knowing More? Mono-Anchored Advantage Normalization for Multi-Source Visual Reasoning

Tsinghua University

Submitted by

taesiri

GenRecon: Bridging Generative Priors for Multi-View 3D Scene Reconstruction

·
5 authors

Submitted by

taesiri

LLMs as Noisy Channels: A Shannon Perspective on Model Capacity and Scaling Laws

·
8 authors

Submitted by

gonzmart

Lost in Sampling: Assessing Lexical Reachability in LLMs via the Word Coverage Score (WCS)

·
6 authors

Submitted by

josephziheng

Less is More: Early Stopping Rollout for On-Policy Distillation

UCLA

University of California, Los Angeles