NeurIPS 2024 First Author Oral · Top 1.8% accepted

Unlocking the Capabilities of Thought: A Reasoning Boundary Framework to Quantify and Optimize Chain-of-Thought

Qiguang Chen, Libo Qin, Jiaqi Wang, Jinxuan Zhou, Wanxiang Che

A reasoning boundary framework that quantifies how far a model's chain-of-thought can go and optimizes it — turning an open-ended ability into a measurable quantity.

Reasoning Boundary Framework: overview of the proposed framework, its combination law and the resulting categories of reasoning boundaries.
SCIS 2026 First Author LMG Best Poster

Towards Reasoning Era: A Survey of Long Chain-of-Thought for Reasoning Large Language Models

Qiguang Chen, Libo Qin, Jinhao Liu, Dengyun Peng, Jiannan Guan, Peng Wang, Mengkang Hu, Yuhang Zhou, Te Gao, Wanxiang Che

A systematic survey of long chain-of-thought reasoning — its landscape, training recipes and evaluation — with an accompanying paper list that has become a widely used reference for the community.

The long chain-of-thought survey teaser: deep reasoning, feasible reflection and extensive exploration as three frontiers of long-CoT research.
SCIS 2026 First Author 950+ papers surveyed

AI4Research: A Survey of Artificial Intelligence for Scientific Research

Qiguang Chen*, Mingda Yang*, Libo Qin, Jinhao Liu, Zheng Yan, Jiannan Guan, Dengyun Peng, Yiyan Ji, Hanjing Li, Mengkang Hu, Yimeng Zhang, Yihao Liang, Yuhang Zhou, Jiaqi Wang, Zhi Chen, Wanxiang Che

A data-centric review of AI for scientific research — organised around tasks, data and evaluation — released together with a living paper collection of 950+ papers.

The AI4Research taxonomy: AI for scientific comprehension, academic survey, discovery, writing and peer review.
ACL 2024 First Author Oral · Top 2.3% submitted

M3CoT: A Novel Benchmark for Multi-Domain Multi-step Multi-modal Chain-of-Thought

Qiguang Chen, Libo Qin, Jin Zhang, Zhi Chen, Xiao Xu, Wanxiang Che

The first benchmark that pushes multi-modal chain-of-thought to multi-domain and multi-step visual reasoning — revealing that VLLMs still answer from textual shortcuts and leave a wide gap to human reasoning.

M3CoT teaser: absence of visual-modal reasoning, single-step visual reasoning and multi-step visual reasoning compared.