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

Aug 2-8

Submitted by

AaronZ345

SwanTale: Unified Multi-Speaker Speech and Audio Generation for Instruct and Zero-Shot Tasks

ByteDance

8

Submitted by

Uanu

DAPD: Dual-Anchored Policy Distillation

Shanghai-AI-Laboratory

Shanghai AI Laboratory

Submitted by

seashell11

LongHorizon-Harness: Advancing Long-Horizon Agents for Real-World Tasks

alibabagroup

Submitted by

KhanCold

MerchantBench: Benchmarking LLM Agents for Long-Term Coherence in E-Commerce Operations

alibabagroup

Submitted by

zli12321

Recursive Synthesis for Long-Horizon Terminal Tasks

Tencent-Hunyuan

Tencent Hunyuan

2

Submitted by

taesiri

JoyAI-Video-Edit: Real-Time Open-Ended Video Editing with Autoregressive Diffusion

jingdong1

Submitted by

zzzzhw

AgentOPSD: Recursive Self-Distillation for Agentic Reinforcement Learning

Tsinghua University

Submitted by

taesiri

Hunyuan3D-Buffalo 1.0: A Unified Multimodal Model for Scalable 3D Generation, Understanding, and Editing

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

Qinsi1

From RLVR to RLSVR: Task Transformation Induces Self-Verifiable Rewards for Open-Ended LLM Self-Improvement

·
11 authors

Submitted by

taesiri

WorldClaw: Agentic 3D Open-World Generation at Scale

Tencent-Hunyuan

Tencent Hunyuan

4

Submitted by

fyv587

AURORA-LM: Autoencoding Unified Representation for Continuous-Latent Diffusion Language Modeling

nanjinguniv

Nanjing University

Submitted by

henggg

N_0-VTLA: Scaling Vision-Tactile-Language-Action Model with Latent Tactile Tokens

NeoteAIEmbodied

Submitted by

isemenkov

Interpretable MEG Decoding of Perceived Speech: Cortical Sources and the Stimulus Features That Drive Retrieval

·
5 authors

Submitted by

PLUS-WAVE

InfiniSplat: Implicit Gaussian Decoding for Large-Baseline Monocular View Synthesis

·
10 authors

Submitted by

QiushiSun

OSReward: Instituting Standardized Evaluation for Cross-Platform Computer-Use Reward Models

hkunlp

NLP Group of The University of Hong Kong

Submitted by

Junlinh

Towards Physics of Multimodal Pretraining: Knowledge Flow, Modality Synergy, Early Unification, and Recipes

Submitted by

scofield7419

Mental World Modeling

mental-world-model

Mental World Model

Submitted by

taesiri

Meshy T2: Fast Native Mesh Generation with Flow Matching

meshy-ai-team

Submitted by

taesiri

ToolArtist: Tool-Using Unified Multimodal Models for Agentic Image Generation

·
8 authors

Submitted by

shenjunhao

Progressive Agent Skill Generation via Reinforcement Learning

CUHKDBGROUP

The Chinese University of Hong Kong - Database Group

Submitted by

hzxllll

Deferred Exposure of Future Trajectories for Verifiable Reasoning in Autonomous Driving VLMs

·
9 authors

Submitted by

CostaliyA

Video-DeepResearch: Towards the Next-Generation Multimodal Deepresearch Agent

·
20 authors

2

Submitted by

songtingyu

UEmbed: Unified Sparse and Dense Multimodal Embeddings

Alibaba-NLP

Submitted by

taesiri

GST-Bench: Can VLMs Develop Global Spatial Awareness from Video?

ByteDance-Seed

Submitted by

FACEONG

N_0-TWAM: Scaling Tactile-Native World-Action Model for Contact-Rich Manipulation

NeoteAIEmbodied

Submitted by

xuzishan

EnvACE: Internalizing Environment Dynamics via World Rehearsal for Agentic Reinforcement Learning

tencent

Submitted by

zhangkangning

VAD: Attributing Visual Evidence for Target Reconstruction in Multimodal On-Policy Distillation

·
12 authors

Submitted by

zhemchuzhnikov

CADENA: Stepwise CAD Reverse Engineering

·
12 authors

Submitted by

heheyas

Scaling Properties of Text Conditioning in Visual Generation

ByteDance-Seed

Submitted by

Kaichengalex

Learning from Failures: Retrieval-Centric CoT via Hard Negatives for Unified Multimodal Retrieval

DeepGlint-AI

Submitted by

SivanSX

ChronoVision: Temporal Reasoning via Latent State Reconstruction

PediaMedAI

2

Submitted by

Yushi98

The Personalization Mirage: How LLMs Fabricate User Profiles, and Why Self-Monitoring Misleads

3

Submitted by

yaya234

PCSD: Persistent Consistency for Self-Distillation in Agentic Reinforcement Learning

BIT

Beijing Institute of Technology

2

Submitted by

taesiri

DiffusionGemma Technical Report

deepmind

Submitted by

ruiye-sjtu

ABSeeker: Training Long-Horizon Search Agents via Answer-Backtracked Credit Assignment

SJTU

Shanghai Jiao Tong University

Submitted by

Ningyu

OneDayAgent: Towards a Long-Horizon Harness for Autonomous Agents

zjunlp

Submitted by

taesiri

HelloWorld: Enabling Socially Interactive Characters in Video World Models

AlayaLab

Submitted by

ldkong

Quo Vadis, World Modeling?

Shanghai-AI-Laboratory

Shanghai AI Laboratory

Submitted by

taesiri

PAST-Bench: Benchmarking the Foundations of Recursive Self-Improvement in Personal Agents

princetonu

Princeton University

Submitted by

Chouoftears

AISPA: User-Centric System Prompt Auditing for Large Language Model Applications

StanfordUniversity

Stanford University

Submitted by

taesiri

HarnessOpt-Bench: Evaluating LLMs at Harness Optimization

ScaleAI

Submitted by

Yuppie1204

WorldExam: Benchmarking World Models from Apparent Appearance to Inherent Reactivity

·
16 authors

Submitted by

bhheo

On-Policy Delta Distillation for Multilingual Math Reasoning

naver-ai

2

Submitted by

Yushi98

Fewer Clarifications, Better Code: Benchmarking Cross-Session Personalized Ambiguity Adaptation in Coding Assistants

3

Submitted by

ayoubkirouane

Teaching Nemotron Greek: Mining a Corpus, Adapting Retrieval, and Grounding Generation for Modern Greek across Specialist Domains

KIEFERSA

3

Submitted by

Benyou

From Economic Agents to Agentic Economies: A Systems Blueprint for Economic World Models

Submitted by

xjcvcvxj

QQWorld: Quantile-Quantile Matching for World Model Regularization

·
3 authors

Submitted by

Artemis0430

SKT: Skill-Use Training at Scale via Verified Synthetic Data Generation

Shanghai-AI-Laboratory

Shanghai AI Laboratory

2

Submitted by

taesiri

K-EXAONE 2.0 Technical Report

LGAI-EXAONE

Submitted by

DogNeverSleep

OmniPack: Unified Token Compression for Efficient Omni-modal Large Language Models

·
13 authors

Submitted by

BugMaker-Boyan

DataSpace: Benchmarking Data Agents for Verifiable Analytics over Heterogeneous Workspaces

HKUSTDial

Submitted by

yyyou

LLaDA MoE v2: Scaling Mixture-of-Experts Diffusion Language Models

GSAI-ML

Submitted by

Sylvest

WCM: A World Critic Model for Vision-Language-Action Reinforcement Learning

OpenMOSS-Team

Submitted by

Yunqiu

SmartMage: Dynamic Modality Orchestration for 3D Scene Understanding

·
5 authors

2

Submitted by

Lingaaaaaaa

Toward Skill-Native LLMs: Skill Entropy for Benchmarking and Training Long-Horizon Reasoning

princetonu

Princeton University

Submitted by

dingyii

SAF-OPD: Stable Advantage Fusion for On-Policy Distillation

meituan-longcat

Submitted by

zd21

Continual Learning in Transition

CASIA

Chinese Academic of Science Institute of Automation

Submitted by

HarrisonPENG

World-to-Wrist: Task-Conditioned Future Wrist Modeling for Fine-Grained Robot Manipulation

·
11 authors

Submitted by

hiyouga

GDPevo: Evaluating Agent Self-Evolution on Real Business Tasks

PrismShadow

Submitted by

wwwyyy

Ego2Robot: Scalable Robot Data Synthesis from Egocentric Human Data

·
15 authors

Submitted by

simingfu

Any-OPD: Heterogeneous On-Policy Distillation for Flow-Matching Models via Representation-Space Bridging

jingdong1

Submitted by

LiuzhipengUCAS

CAPEval: A Decoupled Caption Evaluation across Understanding and Generation

·
3 authors

Submitted by

chestnutlzj

UniWorld-Design: From Pixel Generation to Layer-Native Design

PekingUniversity

Peking University

2

Submitted by

MingxuanXia

Enhancing Rubric-based RL via Self-Distillation

·
10 authors

Submitted by

taesiri

ExtractBench: A Benchmark for Schema-Guided Enterprise Document Extraction

·
5 authors

Submitted by

jenoj

When Teachers Mislead: Spurious-Signal-Aware On-Policy Distillation

ByteDance

Submitted by

Zonghao2025

SkillJack: Persistent Skill Backdoors in Self-Evolving Agents

tencent

Submitted by

Trae1ounG

SWE-Touch: Benchmarking Coding Agents When Users Touch the Code

CASIA

Chinese Academic of Science Institute of Automation

Submitted by

rtyasdf

KVAE: Family of Tokenizers for Multimodal Generative Models

kandinskylab

Submitted by

Longin-Yu

PaDoc: Layout-Grounded Parallel Decoding for Document Parsing

tsinghua-uni

Tsinghua University

Submitted by

GuoxinChen

CalibForge: Adversarial Solver Calibration for Scaling Learnable Terminal Tasks

AweAI-Team

Submitted by

yzx010

GradCuit: Credit-Assigned Gradient Flow Enables Robust and Interpretable Test-Time Latent Reasoning

·
7 authors

Submitted by

Uri-ka

MameLoshnLM: Yiddish Language Model and Evaluation Benchmark

Submitted by

Phoebux

EffectLearner: World-Aware Object-Effect Reasoning for Real-World Video Object Removal

·
11 authors

Submitted by

taesiri

DyPES-VLA: Learning Shared Dynamics Priors and Embodiment-Specific Control for Cross-Embodiment Manipulation

·
15 authors

Submitted by

Dasool

NOLLI: A Difficulty-Calibrated Puzzle Benchmark for Diagnosing the English-Korean Performance Gap

HAERAE-HUB

Submitted by

Jiarui-Wang

MemSFT: Mitigating Alignment Tax with an External Parametric Memory

·
11 authors

Submitted by

gopirajbahadur

To Add Is Machine, To Delete Is Human: Measuring and Mitigating Deletion Avoidance in LLM Code Editing

centre-for-swe

Centre for Software Excellence

Submitted by

zmmm

Motion Beyond Morphology: Bootstrapping Cross-Category Motion Transfer from Abstract Motion Representations

KlingTeam

KlingAI Research

Submitted by

zhaoyian01

MiniWorld: Democratizing the Training of Video World Models from Scratch

PekingUniversity

Peking University

Submitted by

ChangleQu

TurnSight: Turn-Level Hindsight Self-Distillation for Tool-Integrated Reasoning

RUC

Renmin University of China

Submitted by

ParadiseYu

Weak-to-Strong On-Policy Distillation

MicrosoftResearch

Microsoft Research

Submitted by

XUANMINGZHANG

Are the Financial Reasoning from LLMs Credible? A Real World Test over Long-Horizon Statements

Qwen

Submitted by

jasperyan

AgentStream: How Well Do Self-Evolving LLM Agents Perform Under Streaming Tasks?

microsoft

Submitted by

Halcyon-Zhang

DeepVoyager-VL: Incentivizing Vision-in-the-Loop Search for Long-Horizon Multimodal Agents

PekingUniversity

Peking University

Submitted by

Nos7

Activity Frames: Deterministic Screen-Activity Compilation for Agent Memory and Replay

·
1 authors

Submitted by

higokri

3DZip: Spatial-Aware Feature Diversity-Guided Token Compression for 3D Question Answering

·
2 authors

Submitted by

z-rx

LeapTalk: Breaking the Latency-Quality Trade-off in Talking Head Generation

Shanghai-Jiao-Tong-University-SAI

Shanghai Jiao Tong University SAI

Submitted by

LZXzju

Distill Where You Fail: Recovering Learning Signals of Negative RL-Groups from Adaptive Teacher Guidance

·
12 authors

2

Submitted by

levondang

Roomer: Reflective Object-Grounded Model Editing and Repair for 3D Indoor Layout Synthesis

cn-scut

South China University of Technology

Submitted by

taesiri

Evaluation-Verification Reward for Consistent Multi-Reference Image Editing

·
8 authors

Submitted by

GhostCai

MASS: Multiplayer World Models with Authoritative Shared State

AlayaLab

Submitted by

stojnvla

Invisible Shortcuts: Why Vision Encoders Know Your Camera

·
5 authors

3

Submitted by

hyf015

GROVE: Growing and Reasoning over Temporally Stratified Memory from Streaming Video Experience

·
10 authors

Submitted by

Yushi98

When Memory Lies: An Empirical Study of Spatial Memory Staleness in VLM Agents

tencent

2

Submitted by

SivanSX

Decoding Children's Gait Behavior

PediaMedAI

Submitted by

Blancokkk

ContinualSkillBench: Can LLM Agents Truly Evolve Their Capabilities?

PekingUniversity

Peking University

Submitted by

dts1347

DreamTraj: Generating 6-DoF Object Trajectories by Reading Unrendered Video Diffusion Latents

SUSTech

Southern university of science and technology

2

Submitted by

Yuheng02

GaussianSelector: Lightweight Human-Guided Object Selection in 3D Gaussian Splatting with Graph Optimization

lllab

2

Submitted by

XuGuo699

ContextMaster: Interactive Multi-Shot Video Creation via Fixed-Budget Sparse Context Routing

KlingTeam

KlingAI Research

2

Submitted by

mistletoe111

AVE-Compass: Towards Holistic Evaluation for Audio-Video Editing Abilities

NJU-LINK

Submitted by

MarcusGu

WorldCycle: Self-Verifiable Reinforcement Learning for Long-Horizon Video World Models

tencent

2

Submitted by

taesiri

OPD-V: Visual On-Policy Self-Distillation with Modality Balance

·
8 authors

Submitted by

imamnurby

ExplainBench: Evaluating Code Explanations from Agents

·
4 authors

Submitted by

ldxxx

ODEWorld: A Continuous Predictive Architecture via Physical-Time Flow

·
8 authors