Inferring Latent Intentions: Attributional Natural Language Inference in LLM Agents

2026-01-13

AI Summary

Traditional natural language inference (NLI) struggles to model the latent intentions underlying agents’ behaviors in multi-agent environments. To address this limitation, this work proposes Attributional NLI (Att-NLI), a novel framework that introduces abductive-deductive reasoning into NLI for the first time: it generates intention hypotheses through abduction and validates them deductively by incorporating principles from social psychology, thereby establishing a new paradigm for intention attribution. The framework further integrates neuro-symbolic methods by combining large language models with an external theorem prover to enhance logical rigor. Experimental results demonstrate that neuro-symbolic Att-NLI agents achieve an average win rate of 17.08% in the Undercover-V text-based game, significantly outperforming baseline approaches.

Technology Category

• Multiagent Systems: Adversarial Agents
• Machine Learning: Neuro-Symbolic Learning
• Natural Language Processing: Safety and Robustness

Application Category

• Semantics and Knowledge: Data modeling to support human-machine intelligence, including LLMs agents, intelligent system behavior, explanations, and user-friendly interactions
• Search and Retrieval-Augmented AI: Agentic search
• Economics, Online Markets and Human Computation: Humans versus LLMs for data annotation and labeling

Abstract

Attributional inference, the ability to predict latent intentions behind observed actions, is a critical yet underexplored capability for large language models (LLMs) operating in multi-agent environments. Traditional natural language inference (NLI), in fact, fails to capture the nuanced, intention-driven reasoning essential for complex interactive systems. To address this gap, we introduce Attributional NLI (Att-NLI), a framework that extends NLI with principles from social psychology to assess an agent's capacity for abductive intentional inference (generating hypotheses about latent intentions), and subsequent deductive verification (drawing valid logical conclusions). We instantiate Att-NLI via a textual game, Undercover-V, experimenting with three types of LLM agents with varying reasoning capabilities and access to external tools: a standard NLI agent using only deductive inference, an Att-NLI agent employing abductive-deductive inference, and a neuro-symbolic Att-NLI agent performing abductive-deductive inference with external theorem provers. Extensive experiments demonstrate a clear hierarchy of attributional inference capabilities, with neuro-symbolic agents consistently outperforming others, achieving an average win rate of 17.08%. Our results underscore the role that Att-NLI can play in developing agents with sophisticated reasoning capabilities, highlighting, at the same time, the potential impact of neuro-symbolic AI in building rational LLM agents acting in multi-agent environments.

Problem

• attributional inference
• latent intentions
• natural language inference
• multi-agent environments
• abductive reasoning

Innovation

• Attributional NLI
• abductive-deductive inference
• neuro-symbolic AI
• latent intention inference
• multi-agent reasoning