Wenhu Chen

Affiliations

University of Waterloo

papers

3,279

total citations

papers (18)

MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI

CVPR 2024arXiv

1,715

citations

MagicBrush: A Manually Annotated Dataset for Instruction-Guided Image Editing

NEURIPS 2023arXiv

501

citations

Subject-driven Text-to-Image Generation via Apprenticeship Learning

NEURIPS 2023arXiv

234

citations

VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning

NEURIPS 2025arXiv

183

citations

Understanding Reasoning Ability of Language Models From the Perspective of Reasoning Paths Aggregation

ICML 2024arXiv

citations

Harnessing Webpage UIs for Text-Rich Visual Understanding

ICLR 2025arXiv

citations

Vamba: Understanding Hour-Long Videos with Hybrid Mamba-Transformers

ICCV 2025arXiv

citations

Local Explanation of Dialogue Response Generation

NEURIPS 2021arXiv

citations

VISTA: Enhancing Long-Duration and High-Resolution Video Understanding by Video Spatiotemporal Augmentation

CVPR 2025arXiv

citations

Counterfactual Maximum Likelihood Estimation for Training Deep Networks

NEURIPS 2021arXiv

citations

Enhancing Human Experience in Human-Agent Collaboration: A Human-Centered Modeling Approach Based on Positive Human Gain

ICLR 2024arXiv

citations

Wenhu Chen

Affiliations

papers (18)

MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI

MagicBrush: A Manually Annotated Dataset for Instruction-Guided Image Editing

Subject-driven Text-to-Image Generation via Apprenticeship Learning

VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning

UniIR: Training and Benchmarking Universal Multimodal Information Retrievers

MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions

General-Reasoner: Advancing LLM Reasoning Across All Domains

Instruct-Imagen: Image Generation with Multi-modal Instruction

Violin: A Large-Scale Dataset for Video-and-Language Inference

MARBLE: Music Audio Representation Benchmark for Universal Evaluation

Reward Guided Latent Consistency Distillation

Understanding Reasoning Ability of Language Models From the Perspective of Reasoning Paths Aggregation

Harnessing Webpage UIs for Text-Rich Visual Understanding

Vamba: Understanding Hour-Long Videos with Hybrid Mamba-Transformers

Local Explanation of Dialogue Response Generation

VISTA: Enhancing Long-Duration and High-Resolution Video Understanding by Video Spatiotemporal Augmentation

Counterfactual Maximum Likelihood Estimation for Training Deep Networks

Enhancing Human Experience in Human-Agent Collaboration: A Human-Centered Modeling Approach Based on Positive Human Gain

papers (18)

MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI

MagicBrush: A Manually Annotated Dataset for Instruction-Guided Image Editing

Subject-driven Text-to-Image Generation via Apprenticeship Learning

VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning

UniIR: Training and Benchmarking Universal Multimodal Information Retrievers

MagicLens: Self-Supervised Image Retrieval with Open-Ended Instructions

General-Reasoner: Advancing LLM Reasoning Across All Domains

Instruct-Imagen: Image Generation with Multi-modal Instruction

Violin: A Large-Scale Dataset for Video-and-Language Inference

MARBLE: Music Audio Representation Benchmark for Universal Evaluation

Reward Guided Latent Consistency Distillation

Understanding Reasoning Ability of Language Models From the Perspective of Reasoning Paths Aggregation

Harnessing Webpage UIs for Text-Rich Visual Understanding

Vamba: Understanding Hour-Long Videos with Hybrid Mamba-Transformers

Local Explanation of Dialogue Response Generation

VISTA: Enhancing Long-Duration and High-Resolution Video Understanding by Video Spatiotemporal Augmentation

Counterfactual Maximum Likelihood Estimation for Training Deep Networks

Enhancing Human Experience in Human-Agent Collaboration: A Human-Centered Modeling Approach Based on Positive Human Gain