From Shortcuts to Reasoning: Robust Post-Training of Theory of Mind with Reinforcement Learning
2026-06-08 • Machine Learning
Machine Learning
AI summaryⓘ
The authors studied how AI models learn Theory of Mind (ToM), which is the ability to understand others' beliefs and intentions. They found that many current tests give models an easy way to guess right without truly understanding, by using shortcuts based on spurious clues. To fix this, they created better tests without shortcuts and developed a new training method called Thinking-RFT, which combines reinforcement learning with explicit reasoning. Their approach improved model performance on ToM tasks, especially for complex reasoning and new situations, and helped models rely on meaningful cues rather than shortcuts.
Theory of MindReinforcement Fine-TuningSupervised Fine-Tuningspurious correlationsreasoning chainsstate trackingbeliefintentionmultimodal taskscounterfactuals
Authors
Jike Zhong, Yuxiang Lai, Ming Li, Yuheng Li, Wuao Liu, Behzad Dariush, Konstantinos Psounis, Shao-Yuan Lo
Abstract
Theory of Mind (ToM) is a must-acquire skill for modern foundation model systems to operate effectively and safely in the real world. Recent works have explored honing ToM via post-training; however, we show that such progress is confounded by a pervasive "shortcut" issue: tasks can reach up to 99% accuracy by simply exploiting spurious causal correlations, leading to a false sense of ToM. Motivated by this, we first develop a framework to systematically examine ToM datasets for shortcuts and provide guidance for future development. We find that questions reducible to pure state tracking, such as "belief," are especially shortcut-prone compared to mind questions, such as "intention," where reasoning beyond tracking is required. Using four shortcut-free datasets across three ToM contexts, we then comprehensively study whether Reinforcement Fine-Tuning with verifiable rewards and explicit reasoning chains, called Thinking-RFT, elevates ToM beyond Supervised Fine-Tuning, or SFT. Our key findings are as follows. First, Thinking-RFT effectively improves ToM in all scenarios, with a 6% improvement over SFT, particularly in complex higher-order reasoning, with a 10% improvement over SFT, and multimodal cases, with a 7% improvement over SFT. It also generalizes notably better to unseen domains and higher-order queries while being more robust to counterfactuals. Second, ToM benefits specifically from the joint effect of reasoning and RL: Thinking-RFT outperforms Non-Thinking-RFT by 7% on average. Third, RFT works by learning to ground its reasoning on anchor cues, such as keywords and state changes, that correspond to causal factors. We believe our study is useful for developing effective and robust ToM post-training datasets and advancing critical ToM capabilities.