About Me
I am Jun Zhang (张俊), currently pursuing an M.Sc. at the Institute of Software, Chinese Academy of Sciences (ISCAS), where I work in the Chinese Information Processing Laboratory under the supervision of Hongyu Lin and Boxi Cao. I received my bachelor’s degree in Artificial Intelligence from Soochow University. My earlier research focused on temporal reasoning and multimodal agentic reinforcement learning; my current work focuses on learning dynamics and understanding how knowledge is acquired, represented, and updated in large language models. I welcome discussions and collaborations on these topics.
Experience
Worked on data quality improvement and replication experiments for o1-style reasoning, including data curation, validation, and reasoning-oriented workflows.
Research Interests
- LLM Learning Dynamics & Knowledge Mechanisms
- Agentic Reasoning & Alignment
Publications
ReasoningLens: Hierarchical Visualization and Diagnostic Auditing for Large Reasoning Models
TL;DR: An open-source framework for hierarchical visualization and diagnostic auditing of long reasoning chains.
arXiv preprint arXiv:2606.23404
OpenThinkIMG: Learning to Think with Images via Visual Tool Reinforcement Learning
TL;DR: An end-to-end framework for training vision-language models to reason adaptively with external visual tools.
arXiv preprint arXiv:2505.08617
ConflictBank: A Benchmark for Evaluating the Influence of Knowledge Conflicts in LLM
TL;DR: A large-scale benchmark for studying conflicts within model knowledge and between models and external evidence.
Advances in Neural Information Processing Systems (NeurIPS), 2024
Timo: Towards Better Temporal Reasoning for Language Models
TL;DR: A general framework for improving language models across a broad range of temporal reasoning tasks.
Conference on Language Modeling (COLM), 2024
Living in the Moment: Can Large Language Models Grasp Co-Temporal Reasoning?
TL;DR: CoTempQA, a benchmark for evaluating how language models understand and reason about concurrent events.
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (ACL), 2024
Personal Interests
Outside research, I enjoy running marathons, swimming, and playing badminton.