I am a Ph.D. student at the University of California, Santa Barbara (UCSB). My research focuses on Large Language Models, Autonomous Agents, Reasoning, and Long-Context Algorithms.

Research Philosophy: I strive to vertically integrate high-level agentic capabilities with low-level algorithmic innovations, treating model architecture and agent frameworks as unified means to realize scalable and capable AI.

Previously, I graduated from Shanghai Jiao Tong University (ACM Honors Class).

Publications & Manuscripts

Adaptive Counter-Attack Synthesis for Mitigation of Onchain Exploits

• Authors: Hanzhi Liu, Yanju Chen, Jiaming Shan, Jiaxin Song, Chaofan Shou, Hongbo Wen, Yu Feng
• Status: Submitted to ISSTA 2026
• Description: Present Sentinel, a fully automated run-time defense that converts an in-flight exploit into an optimal counter-attack contract on the fly.

Mid-Segment Editing: Teaching Language Models to Maintain Precise State Through Efficient In-Context Editing

• Authors: Jiaming Shan, Yuzhou Nie, Eric Xin Wang
• Status: Targeting COLM 2026
• Description: Proposed a framework enabling LLMs to perform in-place context editing via unified rewrite operations. Developed GroupedRoPEAttention (Triton kernel) for 2-78x speedup.

Proactive Agent Research Environment (PARE): Simulating Active Users to Evaluate Proactive Assistants

• Authors: Deepak Nathani, Cheng Zhang, Chang Huan, Jiaming Shan, Yinfei Yang, Alkesh Patel, Zhe Gan, William Yang Wang, Michael Saxon, Xin Eric Wang
• Status: Submitted to ICML 2026
• Description: Developed PARE, a framework modeling mobile apps as Finite State Machines (FSMs) for evaluating proactive agents.

Building Constrained Human-AI Cooperation: An Inclusive Embodied Social Intelligence Challenge

• Authors: Weihua Du, Qiushi Lyu, Jiaming Shan, Zhenting Qi, Hongxin Zhang, Sunli Chen, Andi Peng, Tianmin Shu, Kwonjoon Lee, Behzad Dariush, Chuang Gan
• Conference: NeurIPS 2024 (Datasets and Benchmarks Track)

Building Cooperative Embodied Agents Modularly with Language Models

• Authors: Hongxin Zhang*, Weihua Du*, Jiaming Shan, Qinhong Zhou, Yilun Du, Joshua B. Tenenbaum, Tianmin Shu, Chuang Gan
• Conference: ICLR 2024