2026
Findings of ACL
Law in Silico: Simulating Legal Society with LLM-Based Agents
An LLM-agent framework for simulating legal societies through individual and institutional decision-making.
00About Me
Yuxuan Chen
I am a second-year MPhil student in the School of Computing and Data Science at The University of Hong Kong, advised by Prof. Chao Huang. My research focuses on self-improving AI agents, agent memory, and large-language-model-based simulation.
The University of Hong KongHong Kong
01 / Updates
News
- OpenSpace v2, the skill management layer for AI agents, was released.
- Our work Law in Silico was published in Findings of ACL 2026.
- We released OpenSpace v1, enabling AI agents to learn, evolve, and share reusable skills across tasks.
- We released FastCode, a fast and cost-efficient framework for repository-scale code understanding.
- We released AI-Trader, a benchmark and platform for autonomous agents in real-time financial markets.
- Our work on LLM-agent generalization using Concordia was published at the NeurIPS 2025 Datasets and Benchmarks Track.
- Simulating Human-like Daily Activities with Desire-driven Autonomy was accepted at ICLR 2025.
- Our team (BIGAI NLCo) ranked #9 in the NeurIPS 2024 Concordia Contest.
02 / Selected work
Recent publications
2026
arXiv preprint
FastCode: Fast and Cost-Efficient Code Understanding and Reasoning
Structural scouting and cost-aware context construction for repository-scale code reasoning.
2025
arXiv preprint
AI-Trader: Benchmarking Autonomous Agents in Real-Time Financial Markets
A live benchmark for evaluating autonomous LLM agents across stocks and cryptocurrency markets.
03 / Open source
Featured projects
Contributor
OpenSpace
A skill management layer that helps AI agents retrieve, evaluate, share, and continuously improve reusable skills.
Contributor
FastCode
A structural-scouting and cost-aware framework for accurate, token-efficient repository-scale code understanding.
Contributor
AI-Trader
A fully automated platform and live benchmark for autonomous agents operating across real-time financial markets.
04 / Background
Research journey
Previous research
LLM Agents for Human Activity Simulation
Beijing Institute for General Artificial Intelligence (BIGAI)
Worked with Prof. Fangwei Zhong on LLM agents for simulating human-like daily activities.
Previous research
Graph Augmentation Benchmarking
Center for Machine Learning Research, Peking University
During an exchange program, benchmarked graph augmentation methods under the supervision of Prof. Wentao Zhang.