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 2

Submitted by

Lanxingxuan

OneStreamer: Unifying Perception, Memory, and Proactive Response in Streaming Video Interaction

NJU

Nanjing University

Submitted by

jpiskorz

On-Policy or Off-Policy Learning? A Systematic Study of Distillation Dynamics

CambUni

University of Cambridge

Submitted by

EddieYang428

Adaptive Reward Routing: Dynamic Multi-Reward Optimization for Joint Audio-Video Diffusion via Forward-Process RL

tencent

Submitted by

rhfeiyang

Hierarchical Continuous Diffusion Language Models

UIUC-CS

University of Illinois at Urbana-Champaign

Submitted by

IDKTreeDP

Beyond Memory: Harnessing Long-Horizon Agents with Explicit Belief States

alibabagroup

Submitted by

changdae

Sharpening Tax in Post-Training

meta

Submitted by

liushiliushi

Agent Priors-guided Policy Learning

NationalUniversityofSingapore

National University of Singapore

Submitted by

MingLiiii

A Missing Piece for Trustworthy AI Reviewers: From Benchmarking Rhetorical Robustness to SciCore Review

·
7 authors

Submitted by

enrue1893

World Observer: Joint Actor-Observer Generation for Persistent World Modeling

kaist-ai

Submitted by

Lunamos

Transformers Stop Thinking Too Early, and a Tiny LoRA Fixes It

·
3 authors

Submitted by

wangzx1994

ROWBench: Do Video Models Render What the Program Specifies?

AlayaLab

Submitted by

ArsenyIvanov

E-MoE: Enhanced Mixture-of-Experts for Non-Factorized Diffusion Language Models

·
5 authors

2

Submitted by

pshlego

ActiveSaddler: Automated Curriculum Learning for Agent Harness Optimization

microsoft

Submitted by

zhaihaotian

Make Sparse Rewards Count: Density-Aware Reward Aggregation for Multi-Reward RL

kuaishou technology

Submitted by

YangC777

AutoGUIWorld: Image Generators as Visual World Models for GUI Agent

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

allonsy07

Retrieval-Augmented Skill Optimization via Cross-Harness Adaptation

·
11 authors

2

Submitted by

shulin16

EgoTools: Towards Tool-Centric Reasoning in Real-World Egocentric Videos

ropedia-ai

Submitted by

xusirui

InterEvolve: Test-Time Evolution of Reward Programs for Humanoid Loco-Manipulation

UIUC-CS

University of Illinois at Urbana-Champaign

Submitted by

tviskaron

Decentralized Master-Mind: Joint Action Refinement through Iterative Intent Denoising in Multi-Agent Pathfinding

·
6 authors

Submitted by

la0ka1

Scaling and Distilling Text Embeddings for Better Diffusibility

umich

University of Michigan

Submitted by

groundhogLLM

GraphForge: Training Working Agents with Graph-Anchored Workspace Synthesis

ustc-community

University of Science and Technology of China

2

Submitted by

chaoyuli

Video Generation Models: A Survey of Post-Training and Alignment

Arizona-State-University

Arizona State University

Submitted by

RuiyangSi

Fewer Tokens, Better Action: GPT-6 Astra Robot Agents with 14% Higher Success Rate but 65% Fewer Tokens

DAGroup-PKU

Submitted by

b1o1o1m

Beyond the Current Scene: Event-Referential Grasping with Active View Selection

Seoul National University

Submitted by

isminoula

SILSA: Sliding-Window Slice Latents for Topology-Preserving High-Resolution 3D Generation

PLAN-Lab

PLAN Lab @University of Illinois Urbana-Champaign

Submitted by

vveicao

4Director: Controlling Video World Models with Rigid 3D Geometry

stabilityai

Submitted by

Lunamos

Persona Dosing: Calibrated Activation Steering for Graded Trait Control

·
7 authors

Submitted by

sitao

X-Tree: Tokenizing Reusable Experience for Efficient Agent Generalization

UWaterloo

University of Waterloo

Submitted by

gowitheflow

AutoDataBench: A Data-centric Testbed for Accelerating Auto Research

AutoDataBench

Submitted by

neosknight

Decoding Looped Transformers Better for (Almost) Free

apple

Submitted by

OldDelorean

Architect-Ant: Editable Automatic Furnishing of Architectural Floor Plans

kaust-generative-ai

KAUST Center of Excellence in Generative AI

2

Submitted by

owl10

Multimodal Flow: Unified Flow Modeling of Language and Vision in Embedding Spaces

HUST Vision Lab

Submitted by

hubin

CorrGRPO: Correlation-Normalized GRPO for Multi-Reward Learning

Submitted by

gtomitsu

Argo-Bench: Evaluating Data Agents on Enterprise-Scale Workflows

textql

Submitted by

chuxuan

RPTune: Learned Context Curation for LLM Catalog Search

google

1

Submitted by

k-m-irfan

Omni-Embed-Mini: Binding Modalities Without Forgetting via Dense Distillation

MBZUAI

Mohamed Bin Zayed University of Artificial Intelligence

Submitted by

HelenPeng

PhysVista: Benchmarking Physical Intelligence in VLMs via a Perception-Reasoning-Assessment Loop

ustc

university of science and technology of china

Submitted by

xsourse

PixelDense: Dense Prediction as Representation Alignment for Pixel Diffusion

·
12 authors

Submitted by

luisrui

Smaller Models, Better Rejects: Preference Distillation Scaling

Submitted by

patrickamadeus

Ego2Act: Evaluating Goal-Directed Manipulation in Egocentric Video Generation

MBZUAI

Mohamed Bin Zayed University of Artificial Intelligence

Submitted by

SophiaSirko

Where-OPD: Spatially Guided On-Policy Self-Distillation of MLLMs with Synthetic Scenes

valeocorg

Submitted by

Yushi98

When Users Change Their Minds: Measuring and Repairing Intent Drift in LLM Agents

·
4 authors

Submitted by

dingyii

Better Supervision Is Nearby: Neighborhood On-Policy Self-Distillation

meituan-longcat

2

Submitted by

AIGNLAI

Pretrain Once, Route Anywhere: Towards a Foundation Model for LLM Routing

nanjinguniv

Nanjing University

Submitted by

al1219

OpenTumorBoard: A Real-World Benchmark of Multidisciplinary Tumor Board Discussion Trajectories

·
12 authors

Submitted by

zomss

AgSpec: Pushing the Limits of Retrieval-Based Speculative Decoding in Coding Agent Pipelines

·
4 authors

Submitted by

nanamma

Benchmarking and Enhancing Skill-Level Memory for Partially Observable Robotic Manipulation

·
7 authors

Submitted by

Yushi98

Do Audio LLMs Listen Before They Act? Diagnosing Acoustic-Context Gating in Voice Agents

·
3 authors

Submitted by

WHB139426

OmniSeek: Native Tool Integration for Multi-turn Audio-Visual Reasoning

·
8 authors

2

Submitted by

lr10260

Align Then Reason: A Multimodal Lip-Sync Judge for Dubbing

netflix

Submitted by

mboss

SemanTok: Predictable Semantic Tokens for Efficient Autoregressive Video Generation

stabilityai

Submitted by

sum0214

LOCI: Spatial Linear Memory for Streaming World Models

IFM

Institute of Foundation Models

Submitted by

Jungang

VTR-Bench: A Systematic Benchmark for Evaluating Visual Text Rendering in Video Generation

·
11 authors

Submitted by

xypkent

DataMagic: Authoring Data Videos through Declarative Multi-Agent Orchestration

·
6 authors

Submitted by

Luffuly

Memorizon: Training World Models Beyond Their Context Window

MBZUAI-IFM

Submitted by

yoonholee

ScholarCatalyst: A Benchmark for Retrieving Papers That Inspire New Research

·
14 authors

Submitted by

haitong-ma

RLE-Bench: A Qualifying Exam for Coding Agents as Robot Learning Engineers

RLE-Bench

Submitted by

naghamo

Generalization Is Stability, Not Accuracy: Multi-Axis Evaluation of LLMs

Technion

Technion Israel institute of technology

1

Submitted by

GeorgeDrayson

Removing the NEEDLE in the Haystack: Backdoor Removal in LLMs via Weight Orthogonalisation

locailabs

Submitted by

271754echo

Devils in Question Relay: Source-Conditioned Relay Steering to Mitigate Hallucinations in Audio-visual Large Language Models

·
6 authors

Submitted by

naufalso

KaliBench: A Fine-Grained Benchmark for Cybersecurity Tool Use on Kali Linux with Runtime-Free Verifiable Rewards

RISys-Lab

Robust Intelligent Systems Lab

Submitted by

Franck-Dernoncourt

Prompt2Skill: Unsupervised Skill Optimization From Natural Language Instructions

·
5 authors

Submitted by

ethanning

Rules to Tools: Executable Checks for LLM Agents in Scientific Computing

AbhiVakil29112001

Carnegie Mellon University

Submitted by

jiangnanhugo

Explore Broadly, Reason Sharply: Push Small Models toward the Frontier via Sampling

morganstanley

2

Submitted by

wjs0725

Does Native 3D Texture Generation Necessarily Require 3D Assets for Training?

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

Franck-Dernoncourt

Personalized Image Generation with Reasoning and Reflection

·
13 authors

Submitted by

Franck-Dernoncourt

FlexRouter: Learning Complementary Model Sets for Flexible LLM Routing

·
9 authors

Submitted by

Franck-Dernoncourt

Joint and Cross-Modal Video-Audio Generation and Editing: A Unified Formulation and Design Taxonomy

·
24 authors

2

Submitted by

TobyYang7

MemFold: Learning Compact Soft Memory for Long-Context Personalization via On-Policy Optimization

·
9 authors

Submitted by

Hoyant-Su

JevSpawn: Adaptive Agentic Inference through Compositional Action Spaces

·
2 authors

Submitted by

rzdiversity

When Does Correction Become Repair? Mechanistic Auditing of Internal Interventions in Tool-Using LLMs

UniversityofBirmingham

University of Birmingham

Submitted by

yunuyean

Prefill-Free Cross-Family KV Cache Transfer for Heterogeneous Multi-Agent LLMs

UniversityofSouthernCalifornia

University of Southern California

Submitted by

Franck-Dernoncourt

Controlled Decoding Attacks on Black-Box LLMs

·
7 authors

Submitted by

1202kbs

Learning What to Recall: Adaptive Multi-Cue Episodic Memory for World Models

kaist-ai

Submitted by

xuehang

It Takes Workflows to Evolve Better Workflows

·
6 authors

Submitted by

OniReimu

Replacing Large Language Models with Jev Decision Models for Low-Latency Edge Service Orchestration

·
5 authors

Submitted by

jsantillana

Keyword Harnesses Fail Open: A Cheap Diagnostic Ladder for Tool-Use Claims in Small Language Models

·
1 authors

2

Submitted by

besset03

OTRetarget: Joint Robot and Object Motion Retargeting via Optimal Transport

INRIA

Institut National de Recherche en Informatique et en Automatique

2

Submitted by

Sta8is

Latent-Foresight: End-to-End Learning Predictable Representations for Latent World Models

·
3 authors

Submitted by

yshin0917

Before It Fades: Reinforcing Temporal Representations at Inference Time in VideoLLMs

·
4 authors

Submitted by

xuehang

Pay for the Fault, Not the Flow: Label-Free In-Flow Multi-Agent Workflow Optimization

·
7 authors

Submitted by

SteveZeyuZhang

DexPolicy: Scheduled Exploration for Trajectory-Guided Dexterous Manipulation

PekingUniversity

Peking University

Submitted by

ethanning

Predictive Credit: Measuring What Scientific Explanations Add to Experimental Forecasts

CarnegieMellonU

Carnegie Mellon University

Submitted by

llama2thedog

Honeycomb: Constant-Size Scene Memory Representation for Video World Models

·
9 authors