About
I am Chunyi Peng(彭淳毅), a first-year M.S. student at Northeastern University, advised by Zhenghao Liu and Yukun Yan. I'm currently interning at ByteDance Seed.
My current research interests focus on Recrusive Self Improvement (RSI), Multimodal Large Language Models (MLLM) and Agentic RL.
News
Our work 'VisRAG2.0: Mitigating Visual Hallucinations via Evidence-Guided Multi-Image Reasoning in Visual Retrieval-Augmented Generation' is accepted by ACM MM 2026!
Our work 'Lang2Act: Fine-Grained Visual Reasoning through Self-Emergent Linguistic Toolchains' is accepted by ACL2026!
Our work 'Mixture-of-Retrieval Experts for Reasoning-Guided Multimodal Knowledge Exploitation' is accepted by SIGIR2026!
UltraRAG has reached 5,500 stars on GitHub!
Selected Publications
View All →CC-OCR V2: Fine-Grained Attribution of LMM Failures in Real-World Visual Document Understanding
Chunyi Peng, Zhipeng Xu, Shuai Bai, Dayiheng Liu, Others
arXiv preprint arXiv:2605.03903
Mixture-of-Retrieval Experts for Reasoning-Guided Multimodal Knowledge Exploitation
Chunyi Peng†, Zhipeng Xu†, Zhenghao Liu, Yukun Yan, Others
SIGIR
This paper introduces MoRE, a multimodal RAG framework that lets MLLMs dynamically coordinate text, image, and table retrieval experts during reasoning. It further proposes Step-GRPO to train expert routing with fine-grained stepwise feedback, improving both answer accuracy and retrieval efficiency.
VisRAG2.0: Mitigating Visual Hallucinations via Evidence-Guided Multi-Image Reasoning in Visual Retrieval-Augmented Generation
Yubo Sun†, Chunyi Peng†, Yukun Yan, Zhenghao Liu, Others
ACM MM
This paper introduces EVisRAG, a visual retrieval-augmented framework that improves multi-image reasoning by explicitly collecting question-relevant evidence from retrieved images before generating an answer. It also proposes RS-GRPO, a reward-scoped training strategy that strengthens evidence grounding and reduces visual hallucinations in visual question answering.
UltraRAG v2: A Low-Code MCP Framework for Building Complex and Innovative RAG Pipelines
Sen Mei, Haidong Xin, Chunyi Peng, Yukun Yan, Others
OpenBMB
A Low-Code MCP Framework for Building Complex and Innovative Retrieval-Augmented Generation systems.