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

July

Submitted by

yh-wang

Orca: The World is in Your Mind

·
57 authors

Submitted by

taesiri

Harness Handbook: Making Evolving Agent Harnesses Readable,Navigable, and Editable

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

xtma

LongStraw: Long-Context RL Beyond 2M Tokens under a Fixed GPU Budget

mindlab-research

Submitted by

AnitaLeung

The Mirage of Optimizing Training Policies: Monotonic Inference Policies as the Real Objective for LLM Reinforcement Learning

·
12 authors

Submitted by

jt-zhang

Vidu S1: A Real-Time Interactive Video Generation Model

Tsinghua University

Submitted by

lmwang

VideoChat3: Fully Open Video MLLM for Efficient and Generalist Video Understanding

MCG-NJU

Multimedia Computing Group-Nanjing University

Submitted by

c7w

Weak-to-Strong Generalization via Direct On-Policy Distillation

BytedTsinghua-SIA

BytedTsinghua-SIA

Submitted by

Chufeng

Boogu-Image-0.1: Boosting Open-Source Unified Multimodal Understanding and Generation

Boogu

Submitted by

yuntian-deng

Program-as-Weights: A Programming Paradigm for Fuzzy Functions

UWaterloo

University of Waterloo

Submitted by

yxl66666

DOPD: Dual On-policy Distillation

·
16 authors

2

Submitted by

YerbaPage

Dockerless: Environment-Free Program Verifier for Coding Agents

ByteDance

3

Submitted by

ubowang

Function-Aware Fill-in-the-Middle as Mid-Training for Coding Agent Foundation Models

TIGER-Lab

Submitted by

mingchenlin2025

ABot-N1: Toward a General Visual Language Navigation Foundation Model

acvlab

Alibaba AMAP CV Lab

Submitted by

taesiri

RynnWorld-4D: 4D Embodied World Models for Robotic Manipulation

·
6 authors

Submitted by

txy

Ring-Zero: Scaling Zero RL to a Trillion Parameters for Emergent Reasoning

inclusionAI

Submitted by

taesiri

AlayaWorld: Long-Horizon and Playable Video World Generation

·
17 authors

Submitted by

Uanu

Accurate, Interdisciplinary and Transparent Structure-property Understanding with Deep Native Structural Reasoning

Shanghai-AI-Laboratory

Shanghai AI Laboratory

Submitted by

Jinyang23

SEED: Self-Evolving On-Policy Distillation for Agentic Reinforcement Learning

·
11 authors

Submitted by

CongWei1230

Search Beyond What Can Be Taught: Evolving the Knowledge Boundary in Agentic Visual Generation

TIGER-Lab

Submitted by

CheeryLJH

ABot-AgentOS: A General Robotic Agent OS with Lifelong Multi-modal Memory

acvlab

Alibaba AMAP CV Lab

Submitted by

taesiri

Video Generation Models are General-Purpose Vision Learners

GoogleDeepMind

Submitted by

huangrh9

Read It Back: Pretrained MLLMs Are Zero-Shot Reward Models for Text-to-Image Generation

ByteDance-Seed

Submitted by

taesiri

RynnWorld-Teleop: An Action-Conditioned World Model for Digital Teleoperation

·
9 authors

Submitted by

rayleizhu

Hierarchical Sparse Attention Done Right: Toward Infinite Context Modeling

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

tttonyalpha

Does VLA Even Know the Basics? Measuring Commonsense and World Knowledge Retention in Vision-Language-Action Models

·
13 authors

Submitted by

seashell11

BlockPilot: Instance-Adaptive Policy Learning for Diffusion-based Speculative Decoding

·
6 authors

Submitted by

chengtan9907

OmniOpt: Taxonomy, Geometry, and Benchmarking of Modern Optimizers

opendatalab

Submitted by

zli12321

Long-Horizon-Terminal-Bench: Testing the Limits of Agents on Long-Horizon Terminal Tasks with Dense Reward-Based Grading

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

BroAlanTaps

UI-MOPD: Multi-Platform On-Policy Distillation for Continual GUI Agent Learning

·
11 authors

Submitted by

amitsaha

SynthDocBench: Controlled Benchmark for Long-Context Visual Document Understanding

ServiceNow-AI

Submitted by

osanseviero

Gemma 4 Technical Report

google

5

Submitted by

taesiri

PixWorld: Unifying 3D Scene Generation and Reconstruction in Pixel Space

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

geuntaek

Video-Oasis: Rethinking Evaluation of Video Understanding

naver

Submitted by

kpzhang996

AgenticSTS: A Bounded-Memory Testbed for Long-Horizon LLM Agents

AlayaLab

Submitted by

taesiri

Scaling Mixture-of-Experts Video Pretraining for Embodied Intelligence

robbyant

Submitted by

yangyu90

ResearchStudio-Reel: Automate the Last Mile of Research from Paper to Poster, Video, and Blog

microsoft

Submitted by

fistyyyy

ResearchStudio-Idea: An Evidence-Grounded Research-Ideation Skill Suite from ML Conference Outcomes

microsoft

Submitted by

taesiri

SearchOS-V1: Towards Robust Open-Domain Information-Seeking Agent Collaboration

antgroup

Submitted by

libr6

Embodied.cpp: A Portable Inference Runtime of Embodied AI Models on Heterogeneous Robots

SEU-PAISys

PhysicalAI System Group @ SAIL Lab, Southeast University

Submitted by

yxl66666

Dual Latent Memory in Vision-Language-Action Models for Robotic Manipulation

·
9 authors

Submitted by

YunxinLi

KnowAct-GUIClaw: Know Deeply, Act Perfectly, Personal GUI Assistant with Self-Evolving Memory and Skill

HIT-TMG

Submitted by

taesiri

Scalable Visual Pretraining for Language Intelligence

·
16 authors

Submitted by

ahnGeo

Why Can't I Open My Drawer? Mitigating Object-Driven Shortcuts in Zero-Shot Compositional Action Recognition

naver

Submitted by

mynameisyoomimi

Scenes as Objects, Not Primitives: Instance-Structured 3D Tokenization from Unposed Views

·
6 authors

2

Submitted by

taesiri

OvisOCR2 Technical Report

·
13 authors

Submitted by

nstar1125

4D Human-Scene Reconstruction from Low-Overlap Captures

SeoulNatlUniv

Seoul National University

Submitted by

selimkuzucu

DataComp-VLM: Improved Open Datasets for Vision-Language Models

·
36 authors

Submitted by

MikaStars39

EvoPolicyGym: Evaluating Autonomous Policy Evolution in Interactive Environments

·
16 authors

Submitted by

landisen

Morphing into Hybrid Attention Models

ByteDance-Seed

Submitted by

MatanAvitan

MANCE: Manifold Aware Concept Erasure

barilan

Bar-Ilan University

Submitted by

LIQIIIII

BadWAM: When World-Action Models Dream Right but Act Wrong

·
3 authors

Submitted by

shixuanke

Vision as Unified Multimodal Generation

sensenova

Submitted by

AndeyTait

SkillHone: A Harness for Continual Agent Skill Evolution Through Persistent Decision History

tencent

Submitted by

Ningyu

LightMem-Ego: Your AI Memory for Everyday Life

zjunlp

Submitted by

taesiri

Infinite Worlds with Versatile Interactions

robbyant

Submitted by

cherubicxn

Vision Pretraining for Dense Spatial Perception

robbyant

Submitted by

DrewJin0827

Multi-Block Diffusion Language Models

SJTU-DENG-Lab

DENG Lab @ SJTU

Submitted by

passing2961

Evolution Fine-Tuning: Learning to Discover Across 371 Optimization Tasks

minnesotanlp

Submitted by

llwswyn

PerceptionRubrics: Calibrating Multimodal Evaluation to Human Perception

JohnsHopkins

Johns Hopkins University

Submitted by

nielsr

Xiaomi-Robotics-U0: Unified Embodied Synthesis with World Foundation Model

XiaomiRobotics

Xiaomi Robotics

Submitted by

lhhuang

Wan-Streamer v0.2: Higher Resolution, Same Latency

Wan-AI

Submitted by

saurabhcantina

OrbitQuant: Data-Agnostic Quantization for Image and Video Diffusion Transformers

Cantina

2

Submitted by

taesiri

Ideas Have Genomes: Benchmarking Scientific Lineage Reasoning and Lineage-Grounded Idea Generation

SJTU

Shanghai Jiao Tong University

Submitted by

huzican

Bridging Interleaved Multi-Modal Reasoning as a Unified Decision Process

tencent

Submitted by

Xingyu-Zheng

Multi-Resolution Flow Matching: Training-Free Diffusion Acceleration via Staged Sampling

·
7 authors

Submitted by

Tsumugii

AgentCompass: A Unified Evaluation Infrastructure for Agent Capabilities

opencompass

Submitted by

Jeff-Wang

GigaWorld-1: A Roadmap to Build World Models for Robot Policy Evaluation

open-gigaai

Submitted by

curtis-sun

AgenticDataBench: A Comprehensive Benchmark for Data Agents

Tsinghua University

Submitted by

jt-zhang

TurboServe: Serving Streaming Video Generation Efficiently and Economically

Tsinghua University

Submitted by

DogNeverSleep

KeyFrame-Compass: Towards Comprehensive Evaluation of Keyframe-Conditioned Video Generation

KlingTeam

Submitted by

taesiri

DSpark: Confidence-Scheduled Speculative Decoding with Semi-Autoregressive Generation

deepseek-ai

Submitted by

yulunliu

LongE2V: Long-Horizon Event-based Video Reconstruction, Prediction, and Frame Interpolation with Video Diffusion Models

NYCU

National Yang Ming Chiao Tung University

Submitted by

KaiiWuu1993

MetaView: Monocular Novel View Synthesis with Scale-Aware Implicit Geometry Priors

Kwai-Kolors

Kolors Team, Kuaishou Technology

Submitted by

Azily

UniClawBench: A Universal Benchmark for Proactive Agents on Real-World Tasks

hkuhk

The University of Hong Kong

Submitted by

kimyuji

LLM-as-a-Tutor: Policy-Aware Prompt Adaptation for Non-Verifiable RL

kaist-ai

2

Submitted by

yshenaw

SkillOpt-Lite: Better and Faster Agent Self-evolution via One Line of Vibe

lmms-lab

Submitted by

taesiri

WorldDirector: Building Controllable World Simulators with Persistent Dynamic Memory

·
13 authors

Submitted by

vanilla1116

AdvancedMathBench: A Benchmark Suite for Advanced Mathematical Proof Generation and Verification

internlm

Intern Large Models

Submitted by

zhxzpn20000506

Trust Region Policy Distillation

Submitted by

zomss

ELDR: Expert-Locality-Aware Decode Routing for PD-Disaggregated MoE Serving

MicrosoftResearch

Microsoft Research

3

Submitted by

taesiri

Seed2.0 Model Card: Towards Intelligence Frontier for Real-World Complexity

ByteDance-Seed

Submitted by

DogNeverSleep

MultiRef-Compass: Towards Comprehensive Evaluation of Multi-Reference-to-Audio-Video Generation

KlingTeam

Submitted by

donghyunli

KronQ: LLM Quantization via Kronecker-Factored Hessian

UniversityofSouthernCalifornia

University of Southern California

Submitted by

zwq2018

VLA-Corrector: Lightweight Detect-and-Correct Inference for Adaptive Action Horizon

OmniAI-ZJU

Submitted by

chengkunli

Blind-Spots-Bench: Evaluating Blind Spots in Multimodal Models

epfl-ch

Submitted by

tangyunbo

MemSyco-Bench: Benchmarking Sycophancy in Agent Memory

·
8 authors

Submitted by

kpzhang996

From Pixels to States: Rethinking Interactive World Models as Game Engines

AlayaLab

Submitted by

mtilyjason

Multi-Turn Agentic Scientific Literature Search via Workflow Induction

illinois

University of Illinois at Urbana-Champaign

Submitted by

Minbyul

Breaking Failure Cascades: Step-Aware Reinforcement Learning for Medical Multimodal Reasoning

·
8 authors

Submitted by

maverickrzw

UniVR: Thinking in Visual Space for Unified Visual Reasoning

ByteDance

Submitted by

lmquan

Concurrent Image Understanding and Generation: Self-Correcting Coupled Markov Jump Processes

google

Submitted by

taesiri

OpenCoF: Learning to Reason Through Video Generation

ByteDance-Seed

Submitted by

taesiri

Multiplayer Interactive World Models with Representation Autoencoders

·
27 authors

Submitted by

wenzhengzeng

Parallelized Autoregressive Decoding for Omni-Modal Dense Video Captioning

NationalUniversityofSingapore

National University of Singapore

Submitted by

fhlyhv

Multimodal Continuous Reasoning via Asymmetric Mutual Variational Learning

·
10 authors

7

Submitted by

Thunderbolt215215

DataEvolver: Self-Evolving Multi-Agent Data Construction for Text-Rich Image Generation

·
5 authors

Submitted by

johncliu

Reinforcement Learning with Metacognitive Feedback Elicits Faithful Uncertainty Expression in LLMs

google

Submitted by

taesiri

MemLearner: Learning to Query Context memory for Video World Models

·
10 authors

Submitted by

Nova2001

PhotoQuilt: Training-Free Arbitrary-Resolution Photomosaics via Bootstrapped Tiled Denoising

uoft

University of Toronto

Submitted by

taesiri

InternVLA-A1.5: Unifying Understanding, Latent Foresight, and Action for Compositional Generalization

InternRobotics

Intern Robotics

Submitted by

LjHuang

EVA-Client: A Unified Data Collection, Inference, and Deployment Framework for Embodied Policies on Real Robots

·
11 authors

Submitted by

Jeff-Wang

GigaWorld-Policy-0.5: A Faster and Stronger WAM Empowered by AutoResearch

open-gigaai

Submitted by

Ray368

Self-Improvements in Modern Agentic Systems: A Survey

·
12 authors

Submitted by

johncliu

Metacognition in LLMs: Foundations, Progress, and Opportunities

yale-nlp

Submitted by

quickjkee

Registers Matter for Pixel-Space Diffusion Transformers

yresearch

Yandex Research