We perform functional grounding of LLMs' knowledge in BabyAI-Text

Integrating GPT-4 with Vision into a reinforcement learning framework using Pygame and TensorFlow, this project enables an agent to navigate and interact within a simulated environment using advanced visual and textual input processing.

Website with interactive client-side CMA-ES (blackbox optimizer) demos. Reinforcement-learning demos allow users to control RL-trained robots.

Composable DSLs for intent parsing, tool calling, and interactive reasoning in `fifo-*` agents

Multimodal dialogue planning for interactive agents in Minecraft-like worlds.

Unity mini-game for the Interactive Agents and Procedural Generation (ECS7016P) module at Queen Mary, University of London.

This project was created for ECS7016P Interactive Agents and Procedural Generation module.