Tsunehiko Tanaka

I am a full-time Software Engineer at IBM, where I work on agent memory for watsonx Orchestrate. I am also a part-time Adjunct Researcher at Waseda University, where I earned my Doctor of Engineering degree in March 2026.

My research has focused on sequential decision-making, drawing on RL and NLP. Recently, I have become interested in working memory for embodied agents. I explored this in my work on long-horizon robotic manipulation, combining VLMs with task and motion planning (TAMP).

During my PhD, my research focused on game AI. I explored LLM post-training with GRPO to turn written board game rules into executable DSL programs. I also developed the Return-Aligned Decision Transformer (RADT), an offline RL method for controlling agent performance via a target score, with the aim of tuning AI opponents’ difficulty.

Email | Google Scholar | X | Github | Linkedin

Publications

	Evidence-Gated Task and Motion Planning with Vision-Language Models Tsunehiko Tanaka, Matthew Stephenson, Alistair Macvicar, Edgar Simo-Serra arXiv preprint 2026 paper
	Return-Aligned Decision Transformer Tsunehiko Tanaka, Kenshi Abe, Kaito Ariu, Tetsuro Morimura, Edgar Simo-Serra TMLR 2025 J2C Certification (presented at ICML 2026) paper \| code \| poster \| OpenReview
	Grammar and Gameplay-aligned RL for Game Description Generation with LLMs Tsunehiko Tanaka, Edgar Simo-Serra IEEE Conference on Games 2025 paper \| code \| DOI
	Grammar-based Game Description Generation using Large Language Models Tsunehiko Tanaka, Edgar Simo-Serra IEEE Transactions on Games 2024 (Early Access) Highlight Paper paper \| code \| DOI
	DiffG-RL: Leveraging Difference between Environment State and Common Sense Tsunehiko Tanaka, Daiki Kimura, Michiaki Tatsubori Findings of EMNLP 2022 paper \| code \| arXiv
	Commonsense Knowledge from Scene Graphs for Textual Environments Tsunehiko Tanaka, Daiki Kimura, Michiaki Tatsubori AAAI Workshop on Reinforcement Learning in Games 2022 paper \| code \| IBM Research
	Challenges in Explainability and Knowledge Extraction Daiki Kimura, Tsunehiko Tanaka, Michiaki Tatsubori, Asim Munawar Wordplay: When Language Meets Games @ NAACL 2022 paper
	LoL-V2T: Large-Scale Esports Video Description Dataset Tsunehiko Tanaka, Edgar Simo-Serra CVPRW (CVSports) 2021 project \| DOI