Reproduce “ToM2Vec: Continuous Theory-of-Mind Tokens for Multi-Agent Coordination” on OpenResearch

Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.

Universal app for Apple silicon and Intel. Requires macOS 11 or later.

About this paper

Nithin Sivakumaran, Ali Payani, Hyunji Lee, Mohit Bansal, Elias Stengel-Eskin

ToM2Vec trains agents to model partners with latent tokens derived from partner traces. During training, an agent predicts ten tokens from local context; at deployment, it uses predictions to condition reasoning and actions, without access to partners’ thoughts. In Overcooked, it scores 106.7 versus 97.6 for a language-summary baseline across three layouts and 30 seeds, evidence limited to this setup.

Summary by alphaXiv.

Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.