Zhengyuan Yang

OR-PRM: A Process Reward Model for Algorithmic Problem in Operations Research

EdiVal-Agent: An Object-Centric Framework for Automated, Fine-Grained Evaluation of Multi-Turn Editing

3D-CoS: A New 3D Reconstruction Paradigm Based on VLM Code Synthesis

STITCH: Simultaneous Thinking and Talking with Chunked Reasoning for Spoken Language Models

TextAtlas5M: A Large-Scale Dataset for Long and Structured Text Image Generation

InfoAgent: Advancing Autonomous Information‑Seeking Agents

V-MAGE: A Game Evaluation Framework for Assessing Vision-Centric Capabilities in Multimodal Large Language Models

Where do Reasoning Models Make a Difference? Follow the Reasoning Leader for Efficient Decoding

Are Unified Vision-Language Models Necessary: Generalization Across Understanding and Generation

MV-Diffus3R: Refining Multi-View Diffusions for Geometric Coherence 3D Reconstruction

Shanks: Simultaneous Hearing and Thinking for Spoken Language Models

Computer-Use Agents as Judges for Automatic GUI Design

Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark

SlowFast-VGen: Slow-Fast Learning for Action-Driven Long Video Generation

VAGEN: Reinforcing World Model Reasoning for Multi-Turn VLM Agents

ViCrit: A Verifiable Reinforcement Learning Proxy Task for Visual Perception in VLMs

SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement

EditRoom: LLM-parameterized Graph Diffusion for Composable 3D Room Layout Editing

Tuning Timestep-Distilled Diffusion Model Using Pairwise Sample Optimization

Point-RFT: Improving Multimodal Reasoning with Visually Grounded Reinforcement Finetuning

Elevating Visual Perception in Multimodal LLMs with Visual Embedding Distillation

GenXD: Generating Any 3D and 4D Scenes

MMWorld: Towards Multi-discipline Multi-faceted World Model Evaluation in Videos

ReFocus: Visual Editing as a Chain of Thought for Structured Image Understanding

OmniContrast: Vision-Language-Interleaved Contrast from Pixels All at once

MMCOMPOSITION: Revisiting the Compositionality of Pre-trained Vision-Language Models