Language Models use Lookbacks to Track Beliefs

• URL: http://arxiv.org/abs/2505.14685v2
• Date: Thu, 23 Oct 2025 17:59:56 GMT
• Title: Language Models use Lookbacks to Track Beliefs
• Abstract summary: We analyze LMs' ability to reason about characters' beliefs using causal mediation and abstraction.<n>Our work provides insights into belief tracking mechanisms, taking a step toward reverse-engineering ToM reasoning in LMs.
• Score: 53.8367373163602
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: How do language models (LMs) represent characters' beliefs, especially when those beliefs may differ from reality? This question lies at the heart of understanding the Theory of Mind (ToM) capabilities of LMs. We analyze LMs' ability to reason about characters' beliefs using causal mediation and abstraction. We construct a dataset, CausalToM, consisting of simple stories where two characters independently change the state of two objects, potentially unaware of each other's actions. Our investigation uncovers a pervasive algorithmic pattern that we call a lookback mechanism, which enables the LM to recall important information when it becomes necessary. The LM binds each character-object-state triple together by co-locating their reference information, represented as Ordering IDs (OIs), in low-rank subspaces of the state token's residual stream. When asked about a character's beliefs regarding the state of an object, the binding lookback retrieves the correct state OI and then the answer lookback retrieves the corresponding state token. When we introduce text specifying that one character is (not) visible to the other, we find that the LM first generates a visibility ID encoding the relation between the observing and the observed character OIs. In a visibility lookback, this ID is used to retrieve information about the observed character and update the observing character's beliefs. Our work provides insights into belief tracking mechanisms, taking a step toward reverse-engineering ToM reasoning in LMs.

Related papers

• Farther the Shift, Sparser the Representation: Analyzing OOD Mechanisms in LLMs [100.02824137397464]
  We investigate how Large Language Models adapt their internal representations when encountering inputs of increasing difficulty.<n>We reveal a consistent and quantifiable phenomenon: as task difficulty increases, the last hidden states of LLMs become substantially sparser.<n>This sparsity--difficulty relation is observable across diverse models and domains.
  arXiv  Detail & Related papers  (2026-03-03T18:48:15Z)
• Mixing Mechanisms: How Language Models Retrieve Bound Entities In-Context [33.223631694438]
  A key component of in-context reasoning is the ability of language models (LMs) to bind entities for later retrieval.<n>We show that this mechanism generalizes poorly to more complex settings.<n>We develop a causal model combining all three mechanisms that estimates next token distributions with 95% agreement.
  arXiv  Detail & Related papers  (2025-10-07T17:44:30Z)
• (How) Do Language Models Track State? [50.516691979518164]
  Transformer language models (LMs) exhibit behaviors that appear to require tracking the unobserved state of an evolving world.<n>We study state tracking in LMs trained or fine-tuned to compose permutations.<n>We show that LMs consistently learn one of two state tracking mechanisms for this task.
  arXiv  Detail & Related papers  (2025-03-04T18:31:02Z)
• Promote, Suppress, Iterate: How Language Models Answer One-to-Many Factual Queries [21.28307426740275]
  To answer one-to-many factual queries, a language model (LM) must simultaneously recall knowledge and avoid repeating previous answers.<n>Across multiple datasets and models, we identify a promote-then-suppress mechanism: the model first recalls all answers, and then suppresses previously generated ones.
  arXiv  Detail & Related papers  (2025-02-27T19:23:15Z)
• Belief Revision: The Adaptability of Large Language Models Reasoning [63.0281286287648]
  We introduce Belief-R, a new dataset designed to test LMs' belief revision ability when presented with new evidence. Inspired by how humans suppress prior inferences, this task assesses LMs within the newly proposed delta reasoning framework. We evaluate $sim$30 LMs across diverse prompting strategies and found that LMs generally struggle to appropriately revise their beliefs in response to new information.
  arXiv  Detail & Related papers  (2024-06-28T09:09:36Z)
• Brittle Minds, Fixable Activations: Understanding Belief Representations in Language Models [9.318796743761224]
  Despite growing interest in Theory of Mind (ToM) tasks for evaluating language models (LMs), little is known about how LMs internally represent mental states of self and others.<n>We present the first systematic investigation of belief representations in LMs by probing models across different scales, training regimens, and prompts.<n>Our experiments provide evidence that both model size and fine-tuning substantially improve LMs' internal representations of others' beliefs, which are structured - not mere by-products of spurious correlations - yet brittle to prompt variations.
  arXiv  Detail & Related papers  (2024-06-25T12:51:06Z)
• Understanding Information Storage and Transfer in Multi-modal Large Language Models [51.20840103605018]
  We study how Multi-modal Large Language Models process information in a factual visual question answering task. Key findings show that these MLLMs rely on self-attention blocks in much earlier layers for information storage. We introduce MultEdit, a model-editing algorithm that can correct errors and insert new long-tailed information into MLLMs.
  arXiv  Detail & Related papers  (2024-06-06T16:35:36Z)
• Leveraging VLM-Based Pipelines to Annotate 3D Objects [68.51034848207355]
  We propose an alternative algorithm to marginalize over factors such as the viewpoint that affect the VLM's response. Instead of merging text-only responses, we utilize the VLM's joint image-text likelihoods. We show how a VLM-based pipeline can be leveraged to produce reliable annotations for 764K objects from the 764K dataset.
  arXiv  Detail & Related papers  (2023-11-29T17:54:22Z)
• How do Language Models Bind Entities in Context? [31.529253469951062]
  Language models (LMs) must bind entities to their attributes to correctly use in-context information. We identify the binding ID mechanism: a general mechanism for solving the binding problem. Our results uncover interpretable strategies in LMs for representing symbolic knowledge in-context.
  arXiv  Detail & Related papers  (2023-10-26T07:10:31Z)
• Towards a Mechanistic Interpretation of Multi-Step Reasoning Capabilities of Language Models [107.07851578154242]
  Language models (LMs) have strong multi-step (i.e., procedural) reasoning capabilities. It is unclear whether LMs perform tasks by cheating with answers memorized from pretraining corpus, or, via a multi-step reasoning mechanism. We show that MechanisticProbe is able to detect the information of the reasoning tree from the model's attentions for most examples.
  arXiv  Detail & Related papers  (2023-10-23T01:47:29Z)
• Knowledge-Augmented Language Model Verification [68.6099592486075]
  Recent Language Models (LMs) have shown impressive capabilities in generating texts with the knowledge internalized in parameters. We propose to verify the output and the knowledge of the knowledge-augmented LMs with a separate verifier. Our results show that the proposed verifier effectively identifies retrieval and generation errors, allowing LMs to provide more factually correct outputs.
  arXiv  Detail & Related papers  (2023-10-19T15:40:00Z)
• Dissecting Recall of Factual Associations in Auto-Regressive Language Models [41.71388509750695]
  Transformer-based language models (LMs) are known to capture factual knowledge in their parameters. We study how the model aggregates information about the subject and relation to predict the correct attribute. Our findings introduce a comprehensive view of how factual associations are stored and extracted internally in LMs.
  arXiv  Detail & Related papers  (2023-04-28T11:26:17Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.