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 6

Submitted by

vvasilev

Kandinsky 6.0 Video: Foundation Models for Synchronized Video and Audio Generation

kandinskylab

Submitted by

Liancheng2

ALoDLM: Adaptively Looped Diffusion Language Models

amazon

Submitted by

Qing145

Memadapter: Counterfactual Adaptation Against Memory-induced Sycophancy

JLU-Deep

DEEP Group at Jilin University

Submitted by

youngyoon911

In-Distribution Forcing for Long Video Generation at Test Time

SeoulNatlUniv

Seoul National University

Submitted by

Lumos-Jiateng

LMBuild: Evaluating LLM Agents for Generating Buildable and Functional Structures

UIUC-CS

University of Illinois at Urbana-Champaign

Submitted by

harryoh99

Foundations of Proactive Agents: Principles, Technical Layers, and Proactivity-Gym

minnesotanlp

2

Submitted by

luochengzz

Self-Generated Feedback Destabilizes Test-Time Training: A Causal Decomposition of Long-Horizon Adaptation

KAUST

King Abdullah University of Science and Technology

Submitted by

mjuicem

OmniReasoning: Pushing the Limits of Audio-Visual Joint Reasoning

Qwen

Submitted by

yifAI

OSWorld-Pro: Process-based Evaluation for Computer Use Agents

nvidia

2

Submitted by

ofantomas

Optimizing the Optimizer: Language Models Discover Faster Molecular Relaxation

·
7 authors

Submitted by

youganglyu

SearchJev: A Fast and Calibrated System-1 Model for Search Agents

·
9 authors

Submitted by

free001style

Representation-Space MMD for Diffusion Language Models

yresearch

Yandex Research

Submitted by

HuskyDoge

Towards Looped Models Done Right, Part II: Rethinking at Fixed Points

IFM

Institute of Foundation Models

Submitted by

LeonMeon

Data Unlearning via Inverse Distillation

·
6 authors

Submitted by

hyojeongyun

CANOPY: Adaptive-Granularity Evidence Compression for Multimodal RAG

POSTECH

Pohang University of Science and Technology

2

Submitted by

jeff024

ASCENT: Online Test-Time Training of Long-Horizon Agents via Self-Distillation of Verified Experience

·
2 authors

2

Submitted by

vinesmsuic

World Editing: Intervening on Executable Worlds at Increasing Depth

UWaterloo

University of Waterloo

Submitted by

Tomo1916

SEER: Self-Evolving Event Reasoning and Retrieval for Time Series Forecasting

google

Submitted by

Leejh123e

RobotUse: Allocating Computation, Context, and Decisions

kaist-ai

Submitted by

Sarim-Hash

Noise Out, Bias In: Targeted Bias Injection in Diffusion Language Models via Closed-Loop Activation Steering

MBZUAI

Mohamed Bin Zayed University of Artificial Intelligence

Submitted by

MingyuLiu

PerturBot: Breaking Shortcut Priors in Vision-Language-Action Models with Perturbative Training

·
7 authors

Submitted by

Sarim-Hash

Certification of Real Images through Calibrated Content Authentication

MBZUAI

Mohamed Bin Zayed University of Artificial Intelligence

Submitted by

SeonghoonYu

When to Switch: Reliable Action-Chunk Extension for Vision-Language-Action Models

POSTECH

Pohang University of Science and Technology

Submitted by

sophicle

Base Models Can Reason By Taking a Cue From Training Data

MIT

Massachusetts Institute of Technology

Submitted by

Ironieser

Rethinking Long-Video Efficiency: A Joint Allocation Perspective on Frames, Pixels, and Front-End Latency

University of Central Florida

Submitted by

yifAI

LoGRA: Scaling LLM Reinforcement Learning with Low-Rank Gradient Sketches

nvidia

Submitted by

yoki123

Capability-Driven Self-Evolution of Agent Memory

microsoft

3

Submitted by

Sterzhang

Video2Skill: From Streaming Experience to Reusable Embodied Skills

·
9 authors

Submitted by

Sterzhang

ProgressCompass: Embodied Progress Reward Models Are Lost Without the Right Context

·
9 authors

Submitted by

TanmaySah

UndoBench: Separating Task Competence from Recovery Capability in Tool-Using AI Agents

·
4 authors

Submitted by

alexeychernysh

QuantCode Model: Specializing Language Models for Executable Algorithmic Trading Code

·
3 authors

1

Submitted by

mack-williams

RealtimeWAM: One-Step Asynchronous World Action Models

NanyangTechnologicalUniversity

Nanyang Technological University

Submitted by

ris3abh-11

When Does Selection Replace Extraction? A Pre-Registered Test of Agent Memory with a Typed Decision Model

·
2 authors

Submitted by

youali

What Matters for Latent Reasoning with Flow Matching

SamsungResearch

Samsung Research

Submitted by

taesiri

Dynamic Harness Search: Building Multi-Agent Systems Per-Query via Prediction

google

Submitted by

Inhenn

LLM-as-Jev: LLMs Are Already Jev-Style Decision Models -- When and How to Fine-Tune Them

microsoft

Submitted by

pedrocurvo

LiFT: Loop Flow Transformers

VISLab-Amsterdam

Video & Image Sense Lab

Submitted by

Debargha

How to Loop MoE: Flatten the Experts, Untie the Attention

·
9 authors

Submitted by

Fulucheng

From Knowledge Access to Source Learning: Developing Source-Specific Competence

GeorgiaTech

Georgia Institute of Technology

Submitted by

shuoxing

The Missing Primitive: Diagnosing and Repairing Mathematical Reasoning in Large Language Models

TexasAMUniversity

Texas A&M University

Submitted by

yuxuanli

PaLoRA: Paced Low-Rank Adaptation for Continual Learning

·
3 authors

Submitted by

rezaesfandiarpoor

Beyond Semantic Similarity: Performance and Costs of Agentic Retrieval for Complex Tasks

nvidia

Submitted by

chenfeng1271

HLA-WM: Hybrid Linear Attention for Long-Horizon Video World Models

Monash University

Submitted by

gpolitis

What Gradients Add to Text Leakage in Split Language Models, Counted per Token and per Document

Setloop

2

Submitted by

mohammad2012191

Periscope: Extending Frozen Language Models Beyond Their Context Window

KAUST

King Abdullah University of Science and Technology

Submitted by

PeterLauLukCh

Training Numerical Intelligence via Auto-Diagnosis and Skill Discovery

·
2 authors

2

Submitted by

FrancisRing

Prism: Dynamic Sparse Attention for Native 2K Joint Video-Audio Generation Model Training

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

Saidgurbuz

DeskForge: Dense Supervision from Desktop Environments for Computer-Use Agents

docling-project

Submitted by

mapooon

Empirical Variational Autoencoder

·
1 authors

Submitted by

Fulucheng

TextReg: Mitigating Prompt Distributional Overfitting via Regularized Text-Space Optimization

GeorgiaTech

Georgia Institute of Technology

Submitted by

yrshi

PluginRSI: Recursive Improvement of Agent Harnesses with Reusable Plugins

·
8 authors

Submitted by

visity

Collaborative Personalized Preference Alignment for LLMs under Data Deficiency

·
3 authors

Submitted by

ptrviyla

COSMI: COmpositional Synthesis of Multi-object Interactions

UniTuebingen

Eberhard Karls Universität Tübingen

Submitted by

JeonghyeokDo

GeoSET: Generalist Foundation Model for SAR-to-EO Image Translation

KAIST-VIClab

KAIST VIClab (Prof. Munchurl Kim)

Submitted by

OniReimu

SoK: Semantic Decision Engines in Network Control Loops

·
6 authors

Submitted by

lennartcb

Learning to Learn a Language

·
2 authors

Submitted by

SteveZeyuZhang

Code2Games: Enabling Coding Agents for Gaming World Generation

PekingUniversity

Peking University

Submitted by

Eric1231

OmniConfess: Eliciting Token Confessions to Mitigate Omni-Modal Hallucination

·
7 authors

Submitted by

yoelzeldes

Closing the Context Gap: Activation Alignment for Tabular In-Context Learning

Independent Research

Submitted by

arminazizi59

Labels Override Definitions in Jev-Style Typed Decision Models

USC

University of Southern California

Submitted by

Mahdip72

Learning Latent Protein Languages for Autoregressive Generation

·
5 authors

Submitted by

aminurhossain

FairRSFM: A Biome-Aware Benchmark and Debiasing Framework for Remote Sensing Foundation Models

·
5 authors

Submitted by

Kivy

MM-ABC: Towards Generalist Mobile Manipulation via Seeing, Coordinating and Imagining

·
9 authors

Submitted by

OniReimu

Intent Interpretation at RIC Timescales: Jev Decision Models versus Large Language Models in 6G Open RAN

·
5 authors

Submitted by

gurayozgur

Learning Steadily: Accumulating Relative Point Margin Scores for Face Image Quality Assessment

FraunhoferIGD

1

Submitted by

chuzy

OpenRUA: Robot-Use Agents Are Zero-Shot Visuomotor Policies

·
7 authors

Submitted by

saeedabdulmuizz

UnAct: Gradient-Free Unlearning via Targeted Activation Intervention

GAASH-Lab

Submitted by

abaggag

The Numerical Linear Algebra of Large Language Models

QCRI

Qatar Computing Research Institute

Submitted by

seffico

Agentic discovery of blood biomarker from distilled private health records

Harvard

Harvard University

Submitted by

arminazizi59

Sharpen Without Search: On-Policy Distillation of Sequence-Level Power Distribution

USC

University of Southern California

Submitted by

yifeipet

Adaptive Fused Prior Transfer for Controllable Generative Image Compression

SantaClaraUniversity

Santa Clara University

Submitted by

thenaivekid

iADD: Improving Alignment and Diversity in Diffusion Policy Optimization

·
5 authors

Submitted by

Maximilian2001

Adapting prior-data fitted networks for tabular anomaly detection

·
2 authors

Submitted by

Agcs12

MEA: A Reward-Driven Multi-Agent System for Faithful Model Explanations

UVABiology

University Of Virginia

Submitted by

JeonghyeokDo

GeoCR: Learning a Generalist Cloud Removal Prior from Heterogeneous Observations

KAIST-VIClab

KAIST VIClab (Prof. Munchurl Kim)

Submitted by

myshi

CurveCodec 2: Skeleton-agnostic animation compression with a learned entropy model

HKU-CGVU

Submitted by

MrBean2024

Arm-wise Compositional Generalization in Dual-Arm Vision-Language-Action Models

·
13 authors

Submitted by

xusirui

InterMimicGen: Scaling Humanoid Loco-Manipulation through Self-Evolving Motion Imitation

·
11 authors

Submitted by

HongjianZhou

The AI Theorist reveals excitonic structure in α-RuCl_3

Oxford

University of Oxford

Submitted by

thenaivekid

Foresight: planning future perception in streaming VLMs without retraining

·
8 authors

Submitted by

asiimo

DEPICT: Scoring Text-to-Image Alignment by Answer Agreement

swordhealth