Developmental Trajectories of Situation Modeling and Mentalizing in Transformer Language Models

2026-06-26

AI Summary

This study investigates whether large language models genuinely possess theory of mind capabilities and how such abilities evolve during training. Drawing on a developmental psychology framework, the authors systematically track the performance of the Olmo2 and Pythia model families across pretraining and post-training stages on contextual modeling and false-belief tasks. They conduct stress tests incorporating supervised fine-tuning (SFT), direct preference optimization (DPO), and distractors containing non-factive verbs. The findings reveal, for the first time, that models of sufficient scale and training exposure exhibit above-chance belief reasoning only in later training phases. Post-training substantially enhances implicit false-belief understanding, yet their contextual representations remain susceptible to linguistic cues, manifesting intrinsic fragility and inconsistency.

Abstract

Recent work suggests that Large Language Models (LLMs) are sensitive to the belief states of agents described by text, as measured by the false belief task (FBT), yet persistent concerns of construct validity remain. We adopt a **developmental perspective**, tracing the pattern of mental state reasoning behavior -- and likely **preconditions** for this behavior -- across multiple training stages in the Olmo2 and Pythia language model suites. We find that above-chance FBT performance depends both on model size and sufficient training volume, emerges relatively late in pretraining, and is most improved by post-training interventions (SFT, DPO) in the condition most diagnostic of mentalizing (False Belief, Implicit). However, FBT performance is fragile: consistent with past work, the use of non-factive verbs (e.g., thinks) increases false belief attributions even in the True Belief condition. To contextualize these findings, we track the emergence of **situation modeling**: the ability to report on basic factual properties of a described scene. Situation modeling accuracy generally precedes and exceeds FBT accuracy, yet situational representations also prove surprisingly incoherent in certain respects: when asked about the knowledge states of the Antagonist agent -- who always knows the item's true location -- Olmo2 13b is consistently influenced both by the Target agent's knowledge state and the presence of non-factive verbs. Together, these results suggest that larger, sufficiently trained models build partially coherent situation models in a developmentally appropriate sequence, yet display surprising fragility -- highlighting the value of developmental and stress-testing approaches for evaluating LLM capabilities.

Problem

• situation modeling
• mentalizing
• false belief task
• developmental trajectories
• large language models

Innovation

• developmental trajectory
• mentalizing
• situation modeling
• false belief task
• large language models