Probabilistic Modeling of Intentions in Socially Intelligent LLM Agents

TL;DR

This paper introduces a probabilistic intent modeling framework for LLM agents in multi-turn social dialogues, improving their ability to adapt and perform better in complex social interactions.

Contribution

The paper presents a novel probabilistic intent modeling approach that dynamically updates beliefs about partner intentions, enhancing social intelligence in LLM agents.

Findings

Increases overall dialogue scores by 9.0% on SOTOPIA-All.

Achieves 4.1% improvement on SOTOPIA-Hard.

Outperforms baseline models and approaches oracle performance.

Abstract

We present a probabilistic intent modeling framework for large language model (LLM) agents in multi-turn social dialogue. The framework maintains a belief distribution over a partner's latent intentions, initialized from contextual priors and dynamically updated through likelihood estimation after each utterance. The evolving distribution provides additional contextual grounding for the policy, enabling adaptive dialogue strategies under uncertainty. Preliminary experiments in the SOTOPIA environment show consistent improvements: the proposed framework increases the Overall score by 9.0% on SOTOPIA-All and 4.1% on SOTOPIA-Hard compared with the Qwen2.5-7B baseline, and slightly surpasses an oracle agent that directly observes partner intentions. These early results suggest that probabilistic intent modeling can contribute to the development of socially intelligent LLM agents.