Explore Theory of Mind: Program-guided adversarial data generation for theory of mind reasoning
It addresses the need for more robust evaluation of theory of mind in AI, which is crucial for social intelligence, but is incremental as it builds on prior benchmarks.
The paper tackles the problem of evaluating theory of mind in large language models by introducing ExploreToM, a framework for generating diverse and challenging data, which reveals that state-of-the-art models achieve accuracies as low as 0% and 9% on this data, and fine-tuning on it improves accuracy on a classic benchmark by 27 points.
Do large language models (LLMs) have theory of mind? A plethora of papers and benchmarks have been introduced to evaluate if current models have been able to develop this key ability of social intelligence. However, all rely on limited datasets with simple patterns that can potentially lead to problematic blind spots in evaluation and an overestimation of model capabilities. We introduce ExploreToM, the first framework to allow large-scale generation of diverse and challenging theory of mind data for robust training and evaluation. Our approach leverages an A* search over a custom domain-specific language to produce complex story structures and novel, diverse, yet plausible scenarios to stress test the limits of LLMs. Our evaluation reveals that state-of-the-art LLMs, such as Llama-3.1-70B and GPT-4o, show accuracies as low as 0% and 9% on ExploreToM-generated data, highlighting the need for more robust theory of mind evaluation. As our generations are a conceptual superset of prior work, fine-tuning on our data yields a 27-point accuracy improvement on the classic ToMi benchmark (Le et al., 2019). ExploreToM also enables uncovering underlying skills and factors missing for models to show theory of mind, such as unreliable state tracking or data imbalances, which may contribute to models' poor performance on benchmarks.