Portfolio item number 1
Short description of portfolio item number 1
Posts by Collection
portfolio
Portfolio item number 2
Short description of portfolio item number 2
publications
Surgical-lvlm: Learning to adapt large vision-language model for grounded visual question answering in robotic surgery
G Wang, L Bai, WJ Nah, J Wang, Zhaoxi Zhang, Z Chen, J Wu, M Islam, H Liu, et al.
Published in ICLR 2024, 2024
We propose Surgical-LVLM, a framework to adapt large vision-language models for grounded visual question answering in robotic surgery scenarios.
Transferring Knowledge from High-Quality to Low-Quality MRI for Adult Glioma Diagnosis
Y Zhao, L Bai, Zhaoxi Zhang, Y Wu, M Islam, H Ren
Published in MICCAI 2024, 2024
A knowledge transfer framework that leverages high-quality MRI data to improve adult glioma diagnosis from low-quality MRI scans.
A survey of uncertainty estimation in llms: Theory meets practice
HY Huang, Y Yang, Zhaoxi Zhang, S Lee, Y Wu
Published in arXiv preprint arXiv:2410.15326, 2024
A comprehensive survey of uncertainty estimation methods in large language models, bridging theoretical foundations with practical applications.
CSVO: Complementary-Pathway Spatial-Enhanced Visual Odometry for Extreme Environments with Brain-Inspired Vision Sensors
Y Lin, Zhaoxi Zhang, Y Chen, T Wang, R Zhao
Published in 2025 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2025
CSVO proposes a complementary-pathway spatial-enhanced visual odometry system tailored for extreme environments, leveraging brain-inspired vision sensors.
3DM: Distill, Dynamic Drop, and Merge for Debiasing Multi-modal Large Language Models
Zhaoxi Zhang, S Lee, Z Wang, Y Wu
Published in ACL 2025, 2025
We propose 3DM, a framework combining distillation, dynamic dropping, and model merging to debias multi-modal large language models.
No free lunch: Rethinking internal feedback for llm reasoning
Y Zhang, Zhaoxi Zhang, H Guan, Y Cheng, Y Duan, C Wang, Y Wang, S Zheng, et al.
Published in arXiv preprint arXiv:2506.17219, 2025
We revisit the role of internal feedback mechanisms in LLM reasoning, revealing fundamental trade-offs and the absence of universally effective strategies.
One Tool Is Enough: Reinforcement Learning for Repository-Level LLM Agents
Zhaoxi Zhang, Y Duan, Y Zhang, Y Xu, Z Wang, K Liang, Y Li, J Liang, D Xia, et al.
Published in ICML 2026, 2025
We show that a single unified tool combined with reinforcement learning is sufficient for training effective repository-level LLM coding agents.
Population-Evolve: a Parallel Sampling and Evolutionary Method for LLM Math Reasoning
Y Zhang, Y Duan, Zhaoxi Zhang, J He, S Zheng
Published in arXiv preprint arXiv:2512.19081, 2025
Population-Evolve introduces a parallel sampling and evolutionary approach to enhance LLM mathematical reasoning through population-based optimization.
RepoZero: Can LLMs Generate a Code Repository from Scratch?
Zhaoxi Zhang, Yiming Xu, Weikang Li, Jiahui Liang, Yunfang Wu
Published in arXiv preprint arXiv:2605.07122, 2026
RepoZero investigates whether LLMs can generate an entire code repository from scratch, evaluating their capability in repository-level code generation.
talks
teaching
Teaching Assistant
Undergraduate course, School of Computer Science, Beijing Institute of Technology, 2024
Course name: Knowledge Engineering
Teaching Assistant
Undergraduate course, School of EECS, Peking University, 2025
Course name: Introduction to Computational Thinking and Algorithms