Biquan Bie is an independent researcher focusing on the theoretical underpinnings and practical applications of LLMs. He graduated from the Experimental High School Attached to Beijing Normal University (SDSZ) in 2020. He received his Bachelor’s degree from the Department of Artificial Intelligence and Automation at Beijing University of Technology (BJUT) in 2024.

He completed his undergraduate thesis on: Explainable Artificial-Intelligence-Based Legal Reasoning Systems

His research focuses on systematically integrating LLMs with domain-specific knowledge to build verifiable automated applications that address precisely defined domain requirements. Additionally, he explores the cognitive processes within LLMs through the lens of human-compatible machine intelligence, employing psychological theories and explainability techniques to understand their reasoning.

Key Projects & Publications:

2024

Trustworthy Legal Reasoning Agent (Internal Testing):

A collaborative project with CUPL focused on developing a deployable multi-agent framework for explainable and reliable legal reasoning.

2025

Research on Anchoring Effects in Large Language Models (Available on arXiv):

A collaborative study with HKUST(GZ) investigates a typical human-compatible cognitive bias anchoring effect in LLMs both internally and externally.

Multi-Agent Framework for Automated Construction of Domain-Specific Large Language Models (Forthcoming on arXiv):

A joint effort with DLUT to develop an end-to-end multi-agent system for building light-weighted specialized LLMs, automating (1) data synthesis, (2) model training and monitoring and (3) downstream knowledge application.

Long-term Research Goals:

His long-term research goals include exploring the application of Reinforcement Learning for LLMs, leveraging Formal Methods to enhance the robustness and reliability of LLMs, and developing LLM-driven automated productivity applications.