Dillon Z. Chen

I am a PhD student working on creating and understanding computational models of intelligence. My research efforts focus on building autonomous agents that can solve long-horizon problems in a range of complex, dynamic environments.

My most recent work focuses on generalised planning – the task of synthesising programs that solve families of planning problems. I design systems that best leverage structured world models in order to few-shot learn policies from simple demonstrations that can efficiently solve problems exhibiting hundreds of relevant objects and solutions with thousands of relevant actions. My results are driven by theoretical understanding and insights rather than engineering efforts.

My work on generalised planning spans the usage of graph neural networks (AAAI’24) and statistical machine learning (NeurIPS’24) for learning efficient and informative value functions, and large language models (PRL@RLC’25) and knowledge representation and reasoning techniques (AAAI’26) for synthesising interpretable and extremely fast policies.

For more research, check out my publications. To play the dog gacha, refresh the page (8 to collect!).