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 30 - Sep 5

Submitted by

namespace-ERI

Repo-To-Skill: Distilling GitHub Repositories Into AI4AI Skills

BAAI

Beijing Academy of Artificial Intelligence

Submitted by

EmpathYang

StudentSim: Training LLM-based Student Simulators

MicrosoftResearch

Microsoft Research

Submitted by

taesiri

Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving

Qwen

5

Submitted by

yuntian-deng

Compile by Training: Turning Natural-Language Specifications into Local Neural Functions

UWaterloo

University of Waterloo

Submitted by

taesiri

Terminal-Universe: Turning Agent Trajectories into Scalable Terminal Environments

Qwen

Submitted by

mozhu

HarnessDev: Can LLMs Create and Evolve Their Own Agent Harness?

ByteDance-Seed

Submitted by

HaoxingChen

LLaDA-Image: Building Strong Image Generators with Fully Open Training Recipes

inclusionAI

4

Submitted by

mozhu

Aspire: Can Models Self-Evolve from Vague Goals?

ByteDance-Seed

Submitted by

Tuwhy

Does On-Policy Distillation Really Distill? From Noisy Teacher to Self-Improvement

Purdue

Purdue University

Submitted by

spongebob0715

Knowing When Not to Reuse: Conditional Experience Transfer in Autonomous LLM Post-Training

·
6 authors

2

Submitted by

junchao-cuhk

SolarWM: Open Data and Scalable Training for Long-Horizon Video World Models

CUHKSZ

Chinese University of Hong Kong, Shenzhen

Submitted by

ChengsongHuang

Random Attention: Rethinking KV Cache Eviction for Efficient Reasoning

Salesforce

Salesforce AI Research

Submitted by

MinghanQin

Lucida: Parse, Generate, and Place for Composable Real-to-Sim Scene Modeling

ByteDance-Seed

2

Submitted by

yunfeixie

RoboTok: An Internet-Scale Data Engine for Human Demonstration Retrieval and Dexterous Manipulation Learning

Rice-RobotPI-Lab

Rice RobotPI Lab

Submitted by

pb09204048

LatentPress: Context Compression Beyond Text and Vision

·
2 authors

Submitted by

YerbaPage

EarlyEval: Cheaper Agent Evaluation via Early Outcome Prediction

SJTU

Shanghai Jiao Tong University

Submitted by

Hangrui-Xu

DART-SD: Diamond-topology Aware Retrieval and Tuning for Self-Distillation of Multi-Turn Tool-Calling Agents

ByteDance

3

Submitted by

hbx

Rethinking On-Policy Distillation of Large Language Models II: One Training Example

Thinking-Space

Submitted by

Senqiao

Beyond Data Scaling: Representation-Centric Continued Pre-training for Vision-Language-Action Models

StarVLA

Star Vision Language Action Models

Submitted by

CheeryLJH

SMELT: Scaling Laws for Compute-Matched MoE Looped Transformers

ByteDance-Seed

Submitted by

dmayboroda

Why Gated DeltaNet Survives 4-Bit Quantization: NVFP4 W4A4 for the Recurrent Half of a Hybrid 27B LLM

minima-ai

2

Submitted by

itsnamgyu

Language Models Can Control Their Own Attention

kaist-ai

Submitted by

Leo-Dai

It Takes Two to Match: Co-Evolving Generative Retriever with Reinforcement Learning

apple

2

Submitted by

Yiyuan

GenFirst: Generation Before Reconstruction for Stable End-to-End Latent Generative Modeling

ByteDance-Seed

2

Submitted by

KangLiao

Puffin-World: Scaling a Unified Multimodal Model with Native 3D World States

ACERobotics

3

Submitted by

GeminiLight

Agentic Artifact Creation: Systems, Evaluation, Principles, and Opportunities

Hong Kong University of Science and Technology(GuangZhou)

Submitted by

taesiri

UI-Venus-2 Technical Report

antgroup

Submitted by

taesiri

On the Design of Qwen3.8-Next Architecture: Evaluation, Efficiency, and Training Stability

Qwen

Submitted by

Jo1uck

Normalized Low-Rank Adaptation

·
5 authors

Submitted by

fenglinglwb

ZimaBlue: Evolving Generalizable World Action Models through Scalable Video Pre-training

JoyFutureAcademy

Joy Future Academy

Submitted by

manchery

Code as Worlds: Agentic Discovery of Executable World Representations for Physical Reasoning

MirroS-Lab

Submitted by

INV-WZQ

H3-World: Turning Language Understanding into World Control

·
5 authors

Submitted by

linjohnss

Scal3R: Learning Efficient Multi-Relative Pose Query for Scalable Online 3D Reconstruction

NYCU

National Yang Ming Chiao Tung University

Submitted by

ChuGyouk

J-Zero: Unified Challenger--Solver--Judge Co-Evolution from Zero Data

kaist-ai

Submitted by

LZXzju

PaperGym: Rubric-Centered Evolution for Research-Plan Generation

·
9 authors

Submitted by

dotvignesh

StarHarness: Evolving Harnesses with Stratified Search for Enterprise Environments

ServiceNow-AI

2

Submitted by

mozhu

S3Gym: Can LLMs Turn Self-Testing and Self-Judging into Self-Improvement?

ByteDance-Seed

Submitted by

harryleafchen

Puro-2B: Poor Lab's Qwen2-1.5B Trained on RTX 5090 within $5090

thu-pacman

PACMAN Group, Tsinghua University

Submitted by

jiachenlei

DreamX-Creator: Democratizing Native Audio-Video Generation at 2K Resolution

GD-ML

Submitted by

HaeChan0305

WHALE: A Simple Recipe for Joint Harness-Weight Optimization

KRAFTON

Submitted by

kenkaneki

From Production Traffic to Post-Training: Building a Self-Hosted LLM That Covers the Corporate Request Mix

t-tech

3

Submitted by

jueunkim

Hi-Q: Hierarchical Evidence-guided Query Refinement for Multi-Hop Question Answering

POSTECH

Pohang University of Science and Technology

2

Submitted by

zouhar

Last Translation Benchmark

·
244 authors

Submitted by

tonywu71

NeoMME: A Single-Tower Multimodal-Native Multilingual Foundation Encoder for Efficient Fine-Tuning and Inference

Hcompany

2

Submitted by

CokeWang

LoopArena: Benchmarking Models as Runtime Controllers for Loop Engineering

GD-ML

Submitted by

yifAI

Fast Weight Attention for Continual Learning

·
11 authors

Submitted by

qian43

Revisiting Local Context for Long-Horizon Streaming 3D Reconstruction

acvlab

Alibaba AMAP CV Lab

Submitted by

gowitheflow

On the Design Fundamentals of Pixel Text Representation Learning

LCO-Embedding

Submitted by

xinhang0111

LightNav-0: Eliciting VLM Spatial Intelligence for Generalist Embodied Navigation

LightOriginsHQ

Submitted by

tsq2000

CogEvol: Towards Efficient and Reliable Learning Environment Generation

CogEvol

元知进化（北京）科技有限公司

Submitted by

visity

Scaling Large Reasoning Models beyond Human Supervision: A Path toward Superintelligence

·
19 authors

Submitted by

quhongyu123

Beyond Retrieval: Progressive Latent Memory Evolution for Streaming Video Understanding

·
10 authors

Submitted by

leesangoh

Act with Intent: Distilling Behavior Intent for Vision-Language-Action Models

POSTECH

Pohang University of Science and Technology

Submitted by

gyuhyeong-k

RealSWE: A Compositional Evaluation of Coding Agents under Realistic User Requests

skku

Sungkyunkwan University

Submitted by

alexWang2002

Evaluating the Hidden Costs of Personalization in Large Language Models

illinois

University of Illinois at Urbana-Champaign

Submitted by

Riasok

SHAPE of Chain-of-Thought in Math Reasoning

SeoulNatlUniv

Seoul National University

Submitted by

Yixuan-Ding-ZJU

LayerRecall: A State-Conditioned Memory Router for Long-Horizon Consistency in Video Generation

zju

Zhejiang University

Submitted by

Hangrui-Xu

Beyond Visual Similarity: Entity-Aligned Retrieval for Knowledge-Based Visual Question Answering

Tsinghua University

2

Submitted by

Sung10230

Super Library Agent: Joint Generation and Maintenance of Multiple Applications Beyond the Single Codebase

kaist-ai

Submitted by

starry0929

The Missing Temporal Link: Temporal Context Routing for Script-Driven Audio-Video Generation

·
10 authors

2

Submitted by

lhmd

ZipTok3D: High-Fidelity 3D Tokenization with Compact Token Prefixes

zju

Zhejiang University

2

Submitted by

gwkrsrch

Evaluating Multimodal LLMs as Generalist Vision-Language-Action Agents for Drone Control: Commanding, Approaching, Tracking and Searching

naver-hyperclovax

Submitted by

Junrulu

ContextPilot: Teaching Agents for Proactive Context Management via Fine-grained RL

tencent

Submitted by

CodeGoat24

WorldReward: Reward Modeling for Camera-Conditioned World Models

·
16 authors

2

Submitted by

songtingyu

CORE: Improving Compositional Reasoning in MLLM Embedding via Reranker Distillation

Alibaba-NLP

Submitted by

shubhamrgandhi

DRACO: Fine-Grained Credit Assignment with Dynamic Rubrics for Long-Horizon Agent Training

ibm-research

Submitted by

jihyoung

PACE: Towards Surfacing Hidden Conflicts in User Requests

POSTECH

Pohang University of Science and Technology

2

Submitted by

craigwu

Uncovering Understanding-Generation Synergy in Native Unified Multimodal Models: From Representation, Task to System

·
6 authors

Submitted by

dingyii

DiagEvo: Diagnosis-Guided Self-Evolution via Hierarchical Error Memory

meituan-longcat

2

Submitted by

231sm

Learning to Evaluate Before Improving: Automatic Rubric Induction for Automatic Research Agents

zjunlp

Submitted by

taesiri

Environment Evolution for Terminal Agents

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

HakHan

Cliff: Learning Process Rewards from the First Mistake

·
6 authors

Submitted by

Mingzh21

Safin-1: Safety from Within through Memory-Native State Evolution

Shanghai-AI-Laboratory

Shanghai AI Laboratory

Submitted by

jiangnanhugo

Using Grounded Theory for Agent Behavior Analysis at Scale

Purdue

Purdue University

Submitted by

byeongjun-park

FlashRender: Few-Step Generative Rendering via Camera-Controlled Video MeanFlow

everex

Submitted by

taesiri

Harness-of-Harness: Multi-Day Autonomous Software Development with Continual Improvement

·
9 authors

Submitted by

taesiri

Matrix-Game 3.5: Enhancing Real-Time Streaming Interactive World Models with Patch Memory

·
17 authors

Submitted by

Hanoona

Locate Anything in Videos: Rethinking Efficient Generative Spatio-Temporal Video Grounding

MBZUAI

Mohamed Bin Zayed University of Artificial Intelligence

Submitted by

Junrulu

Blind Men and the Elephant: Probing the Epistemic Myopia of LLMs under Long-Tail Divergent Knowledge

tencent

Submitted by

Franck-Dernoncourt

A Glance Is All You Need: Single-Pass Fine-Grained Image Captioning with SimLoss

adobe-research

Submitted by

Frinkleko

E-Commerce Bench: Evaluating LLM Agents on Long-Horizon Autonomous Business Operation

Qwen

Submitted by

AlexiaJM

Sliding-window beats linear attention

microsoft

2

Submitted by

zzliang

VibeVoice-ASR-Streaming Technical Report

microsoft

Submitted by

chenguolin

Ring Forcing: Towards Precise Long-Term Memory for Autoregressive Video Diffusion

·
9 authors

2

Submitted by

DongyangChen

Paint What You See: Benchmarking Dexterous Visual Tool Use in Multimodal Agents

Submitted by

taesiri

Principia: Relational Physics Tests for Video Models

·
4 authors

Submitted by

Leo-Dai

Influence-Directed Distillation: Solving the Diversity Bottleneck in Sampled-Token On-Policy Distillation

Leander001

2

Submitted by

yushaohan

StepGuard: Learning Step-Level Guardrails with Scalable Supervision and Safety-Utility Balancing

AI45Research

Submitted by

Prakh25

Select, Compress, Reinvest: A Controlled Study of Visual-Token Allocation in Long-Video MLLMs

·
1 authors

Submitted by

MrSunshy

DramaChain Bench: An End-to-End Benchmark for Short-Drama Generation

Tencent-Hunyuan

Tencent Hunyuan

2

Submitted by

taesiri

Video Generative Models as Geometry Learner

·
5 authors

Submitted by

rsn243

Percolation Dynamics in Optimization : Variance Cascades and Discrete Scale Invariance

·
2 authors

Submitted by

231sm

EM^2Mem: Event-Centric Multimodal Memory for Large Language Models

zjunlp

Submitted by

pminervini

Chain-of-Thought Faithfulness of Reasoning Models Varies with Where and How Preference Cues Are Delivered

EdinburghNLP

EdinburghNLP - Natural Language Processing Group at the University of Edinburgh

Submitted by

yeon-kyeong

Keep-or-Drop? Adaptive Tokenizer for Compact Video Representation

kakaocorp

2

Submitted by

ahnpersie

Lies We Can See: Joint Verbal and Non-Verbal Deception by VLM Agents in Embodied Social Interactions

SeoulNatlUniv

Seoul National University

Submitted by

gujiuxiang

VeriPhy: Agentic Physical Reasoning for World Model Evaluation and Refinement

·
11 authors

Submitted by

HYUDHKIM

Let Confidence Change, Not the Prediction: Prediction-Preserving Repair for Post-hoc Calibration

·
3 authors

Submitted by

lastdefiance20

PonderPounce: A Pretrained MLLM as an Episode Context Engine for Robot Control

maum-ai

Submitted by

YaroslavPrytula

QCell: Recombining and Aligning Cell Queries for Overlapping Instance Segmentation

·
3 authors

Submitted by

taesiri

Post-Training Language Models for Gold-Medal Performance in Coding Competitions

nvidia

Submitted by

taesiri

PaperBanana-Interact: Scientific Diagram Refinement with Multi-Turn Human Feedback

google

Submitted by

taesiri

Scaffolding Foundation Models into Physical-World Agents Pushes the Frontier of Long-Horizon Navigation

·
10 authors

Submitted by

rzdiversity

Locked at the Entrance, Open Inside: Where RLVR Narrows the Solution Space

UniversityofBirmingham

University of Birmingham

Submitted by

MatteoCargnelutti

Institutional Newspapers Pipeline: Deriving billions of high quality tokens from historical newspapers

institutional

Institutional Data Initiative

2