陳幸妤Hsing-Yu Chen
Ph.D. student at NYCU (CGI Lab) and research student at Academia Sinica RLG Lab. I work on reinforcement learning and game AI — Taiwanese Mahjong, Werewolf, and Wall Go — and enjoy building web systems. Coding, CS, and English notes live here.
A belief-shift benchmark that measures how accusations move an LLM's suspicion in Werewolf — 40 open-weight configurations over 1,224 annotated messages. (EMNLP 2026 Main)
Read →An AlphaZero-based agent that masters WallGo — the 7×7 board game popularized by Netflix's The Devil's Plan — beating professional Go players and revealing key strategies. (CG 2026)
Read →Improving opponent modeling in a DRL-based Taiwanese Mahjong AI with heuristics and a neural discard model. Best Paper @ TCGA 2025; Silver @ ICGA Computer Olympiad 2024.
Read →A sustainable online after-class peer-learning platform using AI and blockchain (matching, knowledge management, points). Co-authored; eLearn 2024, Singapore.
Read →NYCU HPC PaaS Platform
A PaaS inference platform being built on NYCU's HPC cluster — 5 NVIDIA DGX nodes (H100 / H200) scheduled with Slurm, over InfiniBand and a WEKA parallel filesystem. Work in progress at the university's IT Service Center.
ResearchWallZero
AlphaZero-based game AI that masters the board game WallGo — my first-author work (CG 2026). Built on MiniZero, learns purely from self-play, and beat professional Go players by ~1.98× territory.
AwardTownPass Recycling
Award-winning recycling microservice built in a 24-hour Taipei City hackathon (top team of 111 nationwide) and shipped live on the official TownPass (台北通) app — LLM trash recognition + garbage-truck alerts, Express.js backend, serverless CI/CD on GCP.
ContributionMiniZero
Open-source zero-knowledge game AI framework (RLG Lab / CGI Lab). I contributed to parts of it — not the original author.