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 9

Submitted by

JarvisPei

NeoHorse-1: Towards Recursive Self-Improvement via Agentic Post-Training with Routing Harness

TokenRhythm

Submitted by

BoJack

AuK Technical Report: An Open-Source Foundational Model for Speech Generation and Editing

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

taesiri

Omni Interaction Agent Technical Report

Tencent-Hunyuan

Tencent Hunyuan

Submitted by

bsmn0223

Eliciting Weak-to-Strong Generalization with On-Policy Reverse Distillation

kaist-ai

Submitted by

knightnemo

OpenWAM: An Open, Modular Exploration Towards Systematic World-Action Model Pretraining

OpenWAM

Submitted by

StarBurger

DriveZero: End-to-End Driving Beyond Human Demonstrations

·
20 authors

Submitted by

ryancll118

GE-Act 2.0: Pretraining and Scaling a World-Action Model for Robotic Manipulation

agibot-world

Submitted by

toshas

Marigold V2: Revisiting Diffusion Transformers for Monocular Depth Estimation

huawei-bayerlab

HUAWEI Bayer Lab

Submitted by

CharyZeng

Miles v0.1: Production-Level Post-Training

RadixArk

Submitted by

Alicezrzhao

Mask Forcing: Improving Autoregressive Video Diffusion Distillation via Dual-Noise Masking Rollout

·
10 authors

Submitted by

xjran

SceneMosaic: Efficient and Diverse Simulation-Ready Scene Generation via Hybrid Agentic Layout Evolution

hkuhk

The University of Hong Kong

Submitted by

kkt20

BeaconKV: Key-Value Cache Compression Guided by Beacon Queries for Efficient Large Reasoning Model Inference

·
4 authors

Submitted by

Codingchild

Reason Through the Latent! Making Latent Visual Reasoning Necessary

·
3 authors

Submitted by

aboots

Steering Geometry: Validating Human Value Geometry in LLM Steering Space

DeepRCL

Submitted by

ngocbh

Kalman Delta Networks: Uncertainty-aware Associative Memory

YaleUniversity

Yale University

Submitted by

Minoday

CosmoH2G: A Hand-to-Gripper Transfer Dataset and Baseline Method for Object Manipulation with Complex Spatial Movements

·
6 authors

Submitted by

MarkWang

Online Draft Co-Training for Speculative Decoding in Large-Scale, Long-Context RL Post-Training

nvidia

2

Submitted by

YinmingHuang

Agentic Visual Generation: From Generative Models to Agentic Control

Fudan-University

Fudan University

Submitted by

taesiri

Procedural Graphs: Self-Evolving Execution Structures for LLM Agents

google

Submitted by

HongbangYuan

Environments as Scaffold: Enriching Feedback to Bootstrap Self-Evolving Agents in Long-Horizon Tasks

FudanUniversity

Fudan University

Submitted by

zxfan

RoboSPA: Can VLA Models Go Beyond Simple Scenes and Short-Horizon Tasks?

zju

Zhejiang University

Submitted by

aviralchharia

CoVeR: Coverage-Based Token Pruning for Multi-View 3D Reasoning in VLMs

CarnegieMellonU

Carnegie Mellon University

Submitted by

taesiri

TANGO: Humanoid Navigation in Cluttered Environments with a Whole-Body Vision-Language-Action Model

·
7 authors

Submitted by

ayoubkirouane

Measuring Language Transfer in Robot Policies: Adding Greek to a Cosmos3 Vision-Language-Action Policy

KIEFERSA

2

Submitted by

Longxiang-ai

TransNormal-2: Geometry-Grounded Rectified Flow with Edge-Aware Decoding for Precise Normal Estimation

·
3 authors

Submitted by

ManTle

VidaForge: Open Research Infrastructure for Video Pretraining Data Recipes

VidaForge

Submitted by

poofuse

What LLM Trading Agents Actually Do in Production: A Six-Month, Population-Scale Record from Two Fleets

DXRG

Submitted by

JiaMao

Multi-Grid Post-Training for Long-Form Multi-Shot Video Generation

·
11 authors

Submitted by

aashiqmuhamed

MOLE: Detecting Insider Threats in AI Agents

CarnegieMellonU

Carnegie Mellon University

Submitted by

robtacconelli

Cadence: Error-Bounded Lossy Compression of Demand Time Series with a Time-Series Foundation Model

·
1 authors

Submitted by

Benyou

What Did I Just Say? Self-Listening for Full-Duplex Speech Models

Submitted by

SoroushMehraban

SynthGait-19K: A Physically Grounded Synthetic Video Dataset for Gait Parameter Estimation

uoft

University of Toronto

Submitted by

XavierJiezou

Harnessing CLIP and DINO: An Uncertainty-Aware Cascaded Fusion Network for Generalizable Deepfake Image Detection

·
7 authors

Submitted by

jiangnanhugo

SQS: Bayesian DNN Compression through Sparse Quantized Sub-distributions

Purdue

Purdue University

Submitted by

cmhungsteve

ReactVAU: A Slow-Fast Decoupled Framework for Streaming Video Anomaly Understanding

nvidia

Submitted by

mokamoto

Encoded Early, Used Late: Where Transformers Begin to Act on an Inferred Partner's Expertise

·
2 authors

2

Submitted by

yucheng-du

Recognition-Refusal Misalignment in LLMs: Why Models Answer Structurally Unanswerable Questions

UniversityofSouthernCalifornia

University of Southern California

Submitted by

xxang

A*-Thought-V2: Efficient Latent Reasoning via Geometric Dynamics of LLM

·
16 authors

2

Submitted by

cyh2004

StudyBench: Can Self-Evolution Squeeze Textbooks for Olympiad Capability?

TsinghuaNLP

Tsinghua NLP Group

Submitted by

elaine1wan

VDiff-Bench: A Challenging Benchmark for Fine-Grained Image Difference Identification

UCLA

University of California, Los Angeles

Submitted by

vLAR

RelightFormer: Feed-forward Generative Transformer for Multiview Object Relighting

PolyUHK

The Hong Kong Polytechnic University

Submitted by

ZixuanKe

EVOHARNESSBENCH: Can Your Agents Keep Pace with an Evolving Harness?

Salesforce

Salesforce AI Research

Submitted by

tsusetzky

NOAH: Learning the Full Patient Journey. A Longitudinal Multimodal Time-Aware Model for Representation and Forecasting

TUM-AIMED

2

Submitted by

gregfrank

Counter-Swarm Doctrine: Containing Coordinated Agent Intrusions

moltaicorp

Submitted by

NCJ

RenderFormer-V2: Neural Rendering with Heterogeneous Scene Primitives

Stanford

Submitted by

lintaihou

Graph Machine: Towards Better Pretraining via Edges

·
1 authors

Submitted by

vrojkova

MasterControl Seventeen Every Time

MasterControlAIML

Submitted by

costwen

Learning 3D Editing without Paired Supervision via Generative Prior Distillation

·
7 authors