ICC Mem/Gen
When a model fine-tunes to believe X but its pretraining says Y — which wins, and can you steer that without retraining?
MSLab · 4-person team
Research in progress
Projects
Research and engineering projects — from LLM knowledge conflicts to quantitative trading to game development.
Research
Algorithm Trading — TWSE
Systematic daily trading strategies for the Taiwan Stock Exchange: signal generation, backtesting, live paper-trading, and a multi-broker execution dashboard.
MIRLab · 3-person team
E-MoTchi — Emoji ↔ Mandarin NLP
Fine-tuned multilingual LLMs for bidirectional translation between emoji sequences and natural Traditional Mandarin, with a from-scratch dataset built around Taiwan internet culture.
5-person team