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

Oct 8

Submitted by

IPF

Self-Retrospection Distillation: Turning Post-hoc Experiences into Prior Foresight

Salesforce

Salesforce AI Research

Submitted by

SempraETY

UltraText Bench: A Comprehensive Bilingual Benchmark for Evaluating Visual Text Rendering in Image Generation

Westlake-University

Westlake University

Submitted by

AaronHuangWei

Long-WAM: Scaling the Context of World-Action Models

nvidia

Submitted by

lgy0404

nanoMuse: An Open-Source Personal Agent for Every Device You Own

zju

Zhejiang University

Submitted by

IMBALDYY

Recursive Game Creator: An Agentic Product-Level Experience-Oriented Game Harness

hkuhk

The University of Hong Kong

Submitted by

chimaharicox

GRACE: Generation-aware latent compression for efficient video generation

kaist-ai

Submitted by

Felix1023

STEPQuant: When and Where Errors Matter in Delta-Rule Recurrent State Quantization

zju-community

Zhejiang University

Submitted by

jinyuan222

Questioning the Questions: Sustaining Self-Evolution in Reasoning Models

HINT-lab

Huang's INTelligence lab

2

Submitted by

ZihanSu

SGF+: Decoupling Gradient Flows for Autoregressive Video Generation

·
13 authors

Submitted by

ZhenlongYuan

VepAgent: Bridging Causal-Transition via Tool-Augmented Reinforcement Learning for Video Event Prediction

UCSC-VLAA

Submitted by

Wenxuan123

UniWAM: Unified World-Action Model

Submitted by

yimingxu

DecepEval: A Benchmark for Evaluating Deception in LLM Agents

XianJiaotongUniversity

Xi'an Jiaotong University

Submitted by

jy9394

Tetris3D: 3D Scene Generation With Objects That Fit Together

kaist-ai

Submitted by

jinheon

RunningTab: Direct Workspace Interaction with Environment-Side Tabs

kaist-ai

Submitted by

Zeyue7

WorldSonus: Bringing Sound to Worlds

NoizAI

Submitted by

HuggingJin

ReSAIL: Mitigating Collapse in Iterative Agent Self-Distillation

RUC

Renmin University of China

Submitted by

visity

RobotWorld: Benchmarking Multimodal Agents for Robot Use Across Diverse Tasks and Embodiments

·
33 authors

Submitted by

SII-xrliu

Mechanics of Long-Context Hybrid Models Part 1.1: From Hybrid Attention to Hybrid Position

OpenMOSS-Team

Submitted by

DtYXs

Semifactual Credit-Augmented Policy Optimization

Westlake-University

Westlake University

Submitted by

hancui

Gains and Collapse in On-Policy Distillation:A Reinforcement Learning Perspective

·
6 authors

Submitted by

zhaocheng

PhysEvo: Astra Can Act, Let It

·
6 authors

2

Submitted by

yifAI

Recurrent Looped Transformer

princetonu

Princeton University

Submitted by

WaltonFuture

SWE-Game: Can Coding Agents Build the Games We Want?

·
11 authors

2

Submitted by

vinesmsuic

VIEScore2: Unified Image Evaluation with Spatially Grounded Explanations

TIGER-Lab

Submitted by

phanviethoang1512

MIMESIS: Learning User Simulators as Training Environments for Interactive Agents

Submitted by

Ryenhails

Inverting Multi-Vector Visual Document Indices

·
3 authors

2

Submitted by

rusrakhimov

NAMVIS: Next-Scale Autoregressive Multi-View Image Synthesis

·
6 authors

Submitted by

ethanning

Agentic RAG Evaluation: Budget Allocation Across Questions, Trajectories, and Reads

CarnegieMellonU

Carnegie Mellon University

Submitted by

yifAI

On KL-Regularized Policy Optimization

math-ai

Submitted by

yonghoon96

Q-Learning with Scalar Adjoint Matching

kaist-ai

Submitted by

tytsui

Minimal Witness Reinforcement Learning

upenn

University of Pennsylvania

Submitted by

j-jaehui

On-Policy Distillation with Negative-Policy Rollouts

naver-ai

Submitted by

yangzhifei

AdSpark: A Large-Scale Dataset and Benchmark for Product-Centric Advertisement Video Generation

PKU

Peking University

2

Submitted by

xiaobiaodu

Mobile-4DGS: Unified Static-Dynamic Real-time Mobile Gaussian Splatting

uts-research

University of Technology Sydney

Submitted by

jianganghan

WebFovea: When the Model Is Right but the Click Is Wrong -- Reliable Round Trips for Vision-Based Web Agents on Live Websites

·
1 authors

Submitted by

bitwxl2022

From Pareto to Preference: Personalized Test-Time Scaling via Amortized Agentic Policy Discovery

BIT

Beijing Institute of Technology

Submitted by

YanSong97

A self-learning scientific agent for X-ray diffraction

Yangtze-ailab

Submitted by

matthewmao

QuadTok: Quadtree Visual Tokenizer for Autoregressive Image Generation

·
7 authors

Submitted by

Geonmo

DLoop: Looped Speculative Decoding

naver-ai

Submitted by

zhemchuzhnikov

CADFather: Autonomous CAD Reconstruction through Coordinated Tool Use

·
10 authors

Submitted by

Yummytanmo

Internalizing Agent Experience into Diffusion Model Weights via On-Policy Context Distillation

·
5 authors

Submitted by

pb09204048

Composing What Each Teacher Learned: Multi-Teacher On-Policy Distillation through Teacher-Relative Shifts

Submitted by

zrchen03

Learning Multimodal Embeddings with Evidence-Aligned Readout

·
11 authors

Submitted by

jinsuby

Improving Proactive AI Assistance with Hierarchical Procedural Understanding

skku

Sungkyunkwan University

3

Submitted by

chgara

Iris-3B: Going Beyond the Latent with Pixel-Space Diffusion Training, Conversion and Fine-Tuning

speridlabs

Submitted by

a43992899

SheetSage2: Coherent Lead-Sheet Transcription with Synthetic Supervision

m-a-p

Multimodal Art Projection

2

Submitted by

YuyaoGe

SkillForge: Co-Evolving Skills and Agents via Dynamic Skill Lifecycles

ict-cas

Institute of Computing Technology, Chinese Academy of Sciences

2

Submitted by

Kotom1

Inherit-MAS: Test-Time Evolution of Multi-Agent Systems through Workflow and Execution Inheritance

·
4 authors

Submitted by

Okiii

UniSkill: Learning Actor-Aligned Skill Proposals for an Evolving Policy

·
7 authors

Submitted by

HongruCai

EngramEdit: Decoupled Knowledge Updates in LLMs through Conditional Memory

ModalityDance

Submitted by

harman

Agent Plasticity: Measuring Self-Improvement Through Experience

meta

2

Submitted by

ghadinehme

StepCAD: Mesh-to-CAD Code Generation via LLM Policy and Geometry-Guided Search

Massachusetts-Institute-of-Technology

Massachusetts Institute of Technology

Submitted by

Corbenic

Real Long-Term Memory for AI: A 50-Million-Token Window That Is Faster and Cheaper Than Recompute

Corbenci

Submitted by

domiso

Salt++: Context-Aligned Post-Training for Few-Step Streaming Multimodal Generation

·
10 authors

Submitted by

aparajic

AutoResearch at Production Scale: Failure Modes and a Multi-Agent Framework

amazon

2

Submitted by

artemZholus

RoboJEPA: Scaling Robotic Latent World Models

meta

Submitted by

yujiazheng

DSReg: Provably Recovering Individual World Latents without Reconstruction

illinois

University of Illinois at Urbana-Champaign

Submitted by

Coral79

PAMI: Part Anchored Motion for Text to Human-Object Interaction Generation

·
5 authors

Submitted by

zhuoqun-chen

CoDance: Learning Reactive and Compliant Human-Humanoid Interaction from Video

DukeUniversity

Duke University

Submitted by

soujanyaporia

RoboQuest: Generalist Physical Agents that Search, Inspect and Test

declare-lab

Deep Cognition and Language Research (DeCLaRe) Lab

Submitted by

Bochkov

Beyond the Parameter Monolith: Reconstructive Memories, Executable Skills, and Residual Assembly for Language Models

·
1 authors

2

Submitted by

jiaxi233

How corner is a corner case? Percentile control for highway scenario generation

·
8 authors

Submitted by

sangaraju1988

Lineage-Aware Memory Governance: A Derivation-Gated Framework for Privacy-Preserving Column-Level Access Control in Enterprise AI Agents

·
2 authors

Submitted by

dexmac

System Switch: When Should a Fast Decision Model Stop and Think?

·
1 authors

Submitted by

Taylantay

TIDES: Implicit Time-Awareness in Selective State Space Models

ethz

Submitted by

yoojinoh

FastOPD: On-Policy Distillation for Lightweight VLA Deployment

kaist-ai

1

Submitted by

jarmstrong001

Task-Sufficient Contraction: Source Selection for Machine Information Interfaces

EricssonAB

2

Submitted by

yfdeng10

Rethinking World-Action Model for Compositional and In-Context Robotic Manipulation

DAGroup-PKU

Submitted by

bw-park

LongTake: Learning to Sustain Dynamics in Long-Horizon Video Generation

kaist-ai

Submitted by

shubhangb

ReGain: Restoring Subject Fidelity in Personalization on Synthetic Images

UIUC-CS

University of Illinois at Urbana-Champaign

Submitted by

XilunZhang

Co-Evolving Robot Orchestrators and Policies through Deployment

StanfordUniversityy

Stanford University

2