We also examine mental states (beliefs, desires, intentions, emotions, knowledge).

🔹 ToMA prioritizes intentions > emotions (other dimensions remain similar)
🔹 Uses +5.6% more 1st-order belief than bases, even when both are prompted equally for 0th/1st order states.

We analyze 4 scenario types: cooperation, negotiation, persuasion, and conflict.

ToMA outperforms the base under all settings. Its reasoning is more strategic (e.g., compromise, accommodation). Even in failures, ToMA shows more active engagement (e.g., failed persuasion).

ToMA adapts effectively to long conversations, sustaining strategic dialogue. When paired with diverse partners, it improves both its own goal completion and its partners’ success.

ToMA generates latent mental states and utterances optimized for social interaction goals using dialogue simulation signals. On Sotopia, it improves performance by +18.9% with Qwen2.5-3B and +6.9% with Qwen2.5-7B, while remaining competitive with a GPT-5 nano baseline.

Theory of Mind can make LLMs better at dialogue: more strategic, goal-oriented, enabling long-horizon adaptation!

In our new paper, we introduce ToMA, a dialogue lookahead training framework that enables the LLMs to generate mental states that are maximally useful for achieving dialogue goals.🧵👇

We are organizing an NLP workshop in Vancouver on Dec 10. Consider registering if you're here for NeurIPS - it's free and open to everyone who is interested in NLP!
We have great members for invited talks and panel discussions.

More details here: nlp.cs.ubc.ca/future-of-nl...