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 29

Submitted by

a43992899

YuE2: Unifying Symbolic and Audio Music Generation at Frontier Quality

m-a-p

Multimodal Art Projection

Submitted by

XINLI1997

Beyond Teacher Assignment: Domain-Normalized Multi-Teacher On-Policy Distillation

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

ZhishanQ

TraceDance: An Automated System for Building Agent Behavior Benchmarks from Real-World Agent Deployment Traces

ByteDance

Submitted by

HongchengGao

Self-Evolving Coding Agents: From Digital Programs to Physical-World Intelligence

HexaFuture

Submitted by

Lines

Post-Training Leaves Behavioral Shadows on Unrelated Decisions

PekingUniversity

Peking University

Submitted by

youyc22

Improving Test-Time Scaling with Adaptive Looped Transformers

nics-efc

Tsinghua-NICS-EFC

Submitted by

lzLiang

EmbodiedMemory-Bench: Benchmarking Embodied Memory for Long-Horizon Embodied Tasks

OmniAI-ZJU

1

Submitted by

yangxw

CompoWorld: Compositional Environment Scaling for General Agents

AllSpark-Research

AllSpark Research

Submitted by

ChengqianMa

Duplex-MPE: Benchmarking Multi-Party Interaction in Full-Duplex Dialogue

PKU

Peking University

Submitted by

tally0818

Knowing When Thinking Is Not Enough: Teaching Small Reasoning Models to Reason Beyond Their Parametric Knowledge

kaist-ai

Submitted by

Ziqi

Learning Native Reflection in Unified Models with Interleaved Reinforcement Learning

·
8 authors

Submitted by

wgcyeo

Surprising Success, Repeated Failure: Entropy-Guided Credit Assignment for Exploration in LLM Reasoning

kaist-ai

Submitted by

YannQi

How Far Are We from Removing the Visual Encoder? Scaling Laws for Encoder-Free Multimodal Pretraining

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

aejion

WorldPlay2: Extending Real-Time Interactive World Models in Control and Horizon

·
8 authors

Submitted by

YangXiao-nlp

Learning to Learn from Context: Synthetic Training from Perturbed Public Documents

AllSpark-Research

AllSpark Research

Submitted by

Shiweiliuiiiiiii

DepthBench: Measuring How Residual Connections Enable More Computational Depth

Intelligent-Systems

Max Planck Institute for Intelligent Systems

Submitted by

yangxw

Skill2Env: Capability-Oriented Environment Synthesis from Skills for General Agents

AllSpark-Research

AllSpark Research

Submitted by

YangCaoCS

SpatialSpeak: QA-Native Reconstruction with Local and Global Context for Spatial Chain-of-Thought Reasoning

The Hong Kong University of Science and Technology

Submitted by

hbx

Diffusion Reward Models

TsinghuaNLP

Tsinghua NLP Group

Submitted by

EthanTaylor

RoboFoundry: System-as-Policy Evolution for Self-Learning Embodied Agents

·
18 authors

Submitted by

kailinjiang

AdaTutoRank: Learning to Rerank Document Sets via Adaptive Tutoring Optimization for RAG and Deep Research

tencent

Submitted by

Peizhen

REALM: A Coarse-to-Fine Generative Framework for Embodied Reactive Listening

Macquarie University

Submitted by

chenmouxiang

QwenGyre: An Elastic Reinforcement Learning Framework for Training xLong-Horizon Agents

Qwen

Submitted by

wukeming11

WideSWE: Can Coding Agents Coordinate Changes Across Repositories?

·
6 authors

Submitted by

huaXiaKyrie

Just MLPs: Efficient Visual State Reconstruction for Multimodal Language Models

tencent

Submitted by

linlinlinlinsdada

Structured Residual Connectivity Matters for Diffusion Transformers

NationalUniversityofSingapore

National University of Singapore

Submitted by

JingzeShi

MassAlloc Attention: Let Attention Allocate Its Own Compute

BAAI

Beijing Academy of Artificial Intelligence

Submitted by

Harold328

SolveEdit: Benchmarking Visual Problem Solving in Generative Models

·
12 authors

Submitted by

ahmedheakl

Program-Verified Self-Evolution for Vision-Language Models

MBZUAI

Mohamed Bin Zayed University of Artificial Intelligence

Submitted by

kzhao5

Rethinking Training-Inference Mismatch in LLM Reinforcement Learning: Where It Arises and How to Correct It

·
7 authors

Submitted by

quantumfr

Imprint Reader: From Weight-Update Readout to Behavioral Intervention

·
3 authors

Submitted by

leoisufa

Precise Editing and Flexible Referencing for Interactable Worlds

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

Garygedegege

SciGen-Verifier: A Multimodal Reasoner for Explainable Verification in Scientific Image Generation

·
9 authors

Submitted by

cskrren

InfiniHand: Streaming World-Space Hand Motion Estimation from Egocentric Video

Shanghai-AI-Laboratory

Shanghai AI Laboratory

Submitted by

KBlueLeaf

TT-VidT: Decoupling the Temporal Axis for Efficient Motion-Centric Video Pretraining

nvidia

Submitted by

Duke-de-Artois

Fewer Tokens, More Self-Teaching: On-Policy Self-Distillation for Extreme Visual Token Reduction

SJTU

Shanghai Jiao Tong University

Submitted by

cskrren

GeoVerse: World-Consistent Novel View Synthesis in Geometric Latent Space

SJTU

Shanghai Jiao Tong University

Submitted by

hongst

RenderRank: Learning to Rerank Text with Compressed Visual Tokens

nlpai-lab

NLP & AI - Korea University

Submitted by

Sssunset

BaRe-Mem: Bayesian Reliability Memory for Robust and Adaptive Agent Consultation

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

47z

Rethinking Automated Voice Similarity by Shifting from EER to Embedding Geometry

NTU

National Taiwan University

Submitted by

DVA13304

An RL View of OPD: Least Square Policy Distillation for Sample-Efficient LLM Reasoning

nvidia

Submitted by

yikaiwang

In-Flight KV Cache with Clean Anchors for Faster Autoregressive Video Diffusion

·
12 authors

Submitted by

LIQIIIII

Residual Transferability in Neural Image Watermarking

·
3 authors

Submitted by

XINLI1997

Measuring Collapse and Correction in Homogeneous-Panel LLM Debate

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

JingzeShi

CoWindow Attention: Full Causal Coverage Is a Collective Property

BAAI

Beijing Academy of Artificial Intelligence

Submitted by

inkoziev

Change the Product, Keep the Parameters: Associative Algebra Layers for Transformers

·
2 authors

Submitted by

LeoYML

TokenCast: Forecasting Token Consumption During LLM Agent Execution

SoutheastU

Southeast Univeristy

Submitted by

Hyeongju97

WaveFront Decoding: Parallelized Self-Speculative Decoding for Looped Language Models

SNU-VLSI

Seoul National University VLSI Lab

Submitted by

mingju0303

Beyond Timestamps: Decision-Aligned On-Policy Distillation for Long-Horizon Agents

Beihang

Beihang University

Submitted by

Ryenhails

ColNanoVDR: Document-Free Query Distillation for Multi-Vector Visual Document Retrieval via Optimal Transport

nanovdr

Submitted by

Gtynnn

When Do Model Internals Help? Exploring the Role of Representation Engineering in LLM Safety

·
3 authors

Submitted by

1e12Leon

DroneWAM: Efficient World Action Model for Drone Visual Navigation

·
8 authors

Submitted by

hjeon2k

KVCMAS: Efficient KV cache Correction for Shared Context in Multi-Agent Systems

SNU-VLSI

Seoul National University VLSI Lab

Submitted by

hjeon2k

PReCache: Efficient KV Cache Sharing for Multi-LoRA Agents via Low-Rank Precomputation and Neutral Reconstruction

SNU-VLSI

Seoul National University VLSI Lab

Submitted by

yunsaijc

Adaptive Consistency Graph for Long-Horizon Agents

Beihang

Beihang University

Submitted by

yanggangu

SMAT: Simple and Efficient Merge-Aware Training

PolyUHK

The Hong Kong Polytechnic University

Submitted by

kcxain

KernelZero: Co-Evolving Proposer and Coder for Continuously Improved GPU Kernel Generation

·
11 authors

Submitted by

metazlb

On-Policy Self-Distillation for Multi-Turn Image Editing

KAUST

King Abdullah University of Science and Technology

Submitted by

Svard

Draft-KV: Learning Useful Latent Communication Between Language Models

tencent

Submitted by

YanZhanPKU

Why Deterministic PRM Guidance Underperforms in Discrete Diffusion Reasoning

PekingUniversity

Peking University

Submitted by

whfeLingYu

SkillDRE: Dual-Stage Red-Team Evolution of Agent Skills via Pre-Execution and Runtime Feedback

·
5 authors

Submitted by

coderchen01

EvolvingAvatar: Interactive 3D Head Generation That Adapts as Conversations Unfold

HFUT-AI

Hefei University of Technology

Submitted by

chenkangjie1123

VGGT-Diff: Visual Geometry Meets Diffusion for Sparse-View Novel View Synthesis

XPENG-AI

Submitted by

ethanning

ControlScope: Workflow Revision and Reliability in LLM Agents

CarnegieMellonU

Carnegie Mellon University

Submitted by

PierreGtch

What masking geometry works best for EEG foundation models?

huggingface

Submitted by

GautamKishore

Safe Error Correction for Language Models: Frozen-Base Adjustment with Capability Preservation

eulogik

Submitted by

GautamKishore

NanoForecast v0.5: Competitive Time Series Forecasting Through Training Pipeline Optimization

eulogik

Submitted by

DaraV

Not All Objectives Are Born Equal: Priority-Constrained Descent for Hierarchical Multi-Objective Optimization

·
2 authors

Submitted by

dipankarsarkar

When Privacy Moves ML-Mediated Decisions On Device: Information and Incentive Misalignment in Auctions

·
1 authors

Submitted by

dipankarsarkar

How Reproducible Are Evaluation Conclusions? A Self-Audit of LLM-Inferred Prompt Structure

·
1 authors

Submitted by

fromthesky

Training and Inference Dynamics of PLDR-LLMs: Row-Map Collapse, Renormalization, and Predictive Reduction

FromtheskyResearchLabs

Fromthesky Research Labs

Submitted by

tafseer-nayeem

How Does "English (US)" Become the Default? Triangulating Structural Bias Towards American English Across the LLM Pipeline

·
2 authors

Submitted by

Yunhao-Feng

AdaGuard: An Adaptive Guard Model with User-defined Policies

·
7 authors

Submitted by

wyy-code

Routing Drift Alone Does Not Diagnose Failure in Merged MoE LLMs

PolyUHK

The Hong Kong Polytechnic University

Submitted by

EnyiJiang

PyroAdapt: Adapting Wildfire Prediction under Spatial Heterogeneity and Temporal Shift

UIUC-CS

University of Illinois at Urbana-Champaign

Submitted by

zyinghua

Rolling-WAM: World Action Models with Rolling Imagination

USC-PSI-Lab

USC Physical Superintelligence (PSI) Lab