Page Not Found
Page not found. Your pixels are in another canvas.
Sitemap
A list of all the posts and pages found on the site. For you robots out there is an XML version available for digesting as well.
Pages
Page not in menu
This is a page not in th emain menu
Posts
Can LLMs Simulate Public Opinion? Initial Findings
Published:
In my current research project, I’ve been exploring whether large language models can accurately simulate how different demographic groups respond to policy questions. Using the Nemotron-Personas-France dataset — which contains 6 million synthetic French adult personas based on the 2025 French census — we’re comparing LLM-generated responses against real public opinion survey data.
Building a Knowledge Graph for Digital Communication: A Student’s Approach
Published:
Over the past semester, I’ve been building a comprehensive knowledge graph covering 14 areas of digital marketing communication. What started as a simple note-taking exercise evolved into something much more interesting: a structured, queryable map of how concepts in digital communication connect to each other.
portfolio
LLM-Assistant
Published:
AI assistant application with MCP (Model Context Protocol) integration — open source.
Douban Movie Crawler
Published:
Automated web scraping tool for collecting movie information from Douban — open source.
Kaggle NCAA March Madness Prediction
Published:
Predictive modeling competition — Top 40% finish in NCAA basketball tournament bracket prediction.
LLM Population Attitude Simulation
Published:
Research project comparing LLM-simulated policy attitudes against real-world French survey data using the Nemotron-Personas-France dataset.
RouteJudge
Published:
Full-stack application for preference-based evaluation of LLM routers under pluralistic user preferences.
publications
Priors in LLM Routing: From Initialization to Regularization
Preprint, 2026
Investigates how prior knowledge can be incorporated into LLM routing, from weight initialization through regularization.
Recommended citation: Guannan Lai, Haoran Hu, and Han-Jia Ye. (2026). Priors in LLM Routing: From Initialization to Regularization. Preprint. https://openreview.net/pdf?id=K3xNTJOM1j
From Sampled Outcomes to Capability Distributions: Rethinking Supervision for LLM Routing
Published in EMNLP 2026, 2026
Reframes LLM routing supervision from discrete outcome sampling to continuous capability distribution modeling.
Recommended citation: Guannan Lai, Haoran Hu, Long Chen, Zhenguo Li, and Han-Jia Ye. (2026). From Sampled Outcomes to Capability Distributions: Rethinking Supervision for LLM Routing. EMNLP 2026. https://arxiv.org/abs/2606.06924
RouteJudge: Preference-Based Evaluation of LLM Routers under Pluralistic User Preferences
Published in ICML 2026 Workshop, 2026
Proposes a preference-based evaluation framework for LLM routers accounting for diverse user preferences.
Recommended citation: Guannan Lai, Haoran Hu, and Han-Jia Ye. (2026). RouteJudge: Preference-Based Evaluation of LLM Routers under Pluralistic User Preferences. ICML 2026 Workshop. https://openreview.net/forum?id=qx8X5hC9Kq
ORBIT: An Optimal Routing and Budgeted Inference Toolbox
Published in Frontiers of Computer Science, 2026
Introduces a modular, extensible toolbox for standardized LLM routing research and budget-aware evaluation.
Recommended citation: Guannan Lai, Haoran Hu, Hao-Xuan Ma, and Han-Jia Ye. (2026). ORBIT: An Optimal Routing and Budgeted Inference Toolbox. Frontiers of Computer Science. https://www.lamda.nju.edu.cn/laign/file/FCS-261310-final.pdf