2024/05/14 by Akhila Yerukola, Yerukola, Akhila, Saujas Vaduguru +5
Arts and Humanities · Business, Management and Accounting · #Artificial Intelligence (cs.AI) #Computation and Language (cs.CL) #FOS: Computer and information sciences #Taxation and Legal Issues #Theology and Canon Law Studies
paper · pdf · doi:10.48550/arxiv.2405.08760
openalex publication_date 2024/05/14 · openalex created_date 2024/05/16 · openalex updated_date 2026/07/28
Humans often express their communicative intents indirectly or non-literally, which requires their interlocutors -- human or AI -- to understand beyond the literal meaning of words. While most existing work has focused on discriminative evaluations, we present a new approach to generatively evaluate large language models' (LLMs') intention understanding by examining their responses to non-literal utterances. Ideally, an LLM should respond in line with the true intention of a non-literal utterance, not its literal interpretation. Our findings show that LLMs struggle to generate pragmatically relevant responses to non-literal language, achieving only 50-55% accuracy on average. While explicitly providing oracle intentions significantly improves performance (e.g., 75% for Mistral-Instruct), this still indicates challenges in leveraging given intentions to produce appropriate responses. Using chain-of-thought to make models spell out intentions yields much smaller gains (60% for Mistral-Instruct). These findings suggest that LLMs are not yet effective pragmatic interlocutors, highlighting the need for better approaches for modeling intentions and utilizing them for pragmatic generation.