Model improves conversation understanding by using social context cues
CLUE: Uncovering Social Context for Pragmatic Understanding
Computation and Language
Summary
Understanding conversations means knowing not just what people say but also their relationships and earlier chat. The authors created a method called CLUE that helps AI models find and use clues from the conversation itself to better understand social meanings. CLUE learns to explain its reasoning clearly, using only the conversation as evidence. When tested, it performed better than a strong baseline model, requiring less thinking time and matching human judgments more closely.
What this means in practice
- •For customer support teams: Enhance AI chatbots to better grasp customer relationships and previous interactions for more accurate responses.
- •For social media analysts: Analyze conversations with improved context awareness to uncover social dynamics and intentions more reliably.
Authors
Chengfei Wu, Zhaoqing Wu, Dan Goldwasser
Abstract
Understanding a conversation requires connecting what is said to the speakers' relationships, expectations, and earlier exchanges. Prior work shows that supplying relevant context can improve models' social judgments, but much of this information may already be present in the conversation. This motivates training models to discover the cues that matter and use them to support an interpretation. We propose CLUE, a two-stage training pipeline that combines supervised evidence reasoning with GRPO guided by Grounded Information Utilization ($\mathrm{IU}_{\mathrm{G}}$). This reward credits new, evidence-supported social inferences while accounting for reasoning effort. We train on a small set of conversational examples to assess effectiveness and generalization across tasks, languages, and genres. At inference time, the trained model reasons over the supplied interaction in a self-contained manner. Compared with base Qwen, CLUE achieves relative gains of 17.3% in-domain and 9.3% out-of-domain in average weighted F1, with higher F1 on four of five out-of-domain evaluations. These gains are accompanied by over 80% fewer reasoning tokens on average in both domains. Human evaluation further finds that CLUE more often grounds its explanations in specific conversational content, with closer alignment between its quoted evidence and human-selected context.