Understanding Why Language Models Hallucinate: Testing Reasoning Against Priors

• URL: http://arxiv.org/abs/2607.00447v1
• Date: Wed, 01 Jul 2026 05:02:43 GMT
• Title: Understanding Why Language Models Hallucinate: Testing Reasoning Against Priors
• Abstract summary: Large language models often produce hallucinated answers that violate prompt-level constraints.<n>We study inference misalignment: a mismatch between the answer supported by the prompt and the answer favored by statistically salient latent associations.<n>We formalize this view with a latent key-task model, in which pretraining-frequency imbalance can cause a shortcut path to dominate the constraint-sensitive path.
• Score: 11.702863087019809
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models often produce hallucinated answers that violate prompt-level constraints. A key diagnostic question is whether these failures reflect missing knowledge, or whether the model has the relevant information but follows the wrong inference path. We study this phenomenon as inference misalignment: a mismatch between the answer supported by the prompt and the answer favored by statistically salient latent associations. We formalize this view with a latent key-task model, in which pretraining-frequency imbalance can cause a shortcut path to dominate the constraint-sensitive path and induce positive inference loss. The framework predicts two failure modes: task-retrieval bias in entity disambiguation and key-selection bias in action choice. We introduce TrapQA, a controlled diagnostic testbed with two components. ScientistQA tests disambiguation among similar scientists with supplementary factual probes, while Real-Life Constrained QA tests everyday constraint following under salient shortcuts. Our results show that hallucination can arise from biased latent inference rather than absent knowledge alone.

Related papers

• When the Wrong Key Wins: Understanding and Detecting Hallucinations in LLMs [6.246552320858934]
  Large language models can hallucinate even when the knowledge required for a correct answer is already available.<n>We show that model predictions can be highly sensitive to individual query keywords.<n>We introduce a two-stage keyword-perturbation method for hallucination detection.
  arXiv  Detail & Related papers  (2026-09-14T06:35:28Z)
• What Am I Missing? Question-Answering as Hidden State Probing [13.065059683491958]
  We propose to leverage question-asking as an inference-time intervention.<n>We train a probe on the student's hidden state before and after asking a question.<n>We find that the success of question-asking as an intervention is largely dependent on the model's self-consistency.
  arXiv  Detail & Related papers  (2026-05-29T17:27:07Z)
• Decoding Answers Before Chain-of-Thought: Evidence from Pre-CoT Probes and Activation Steering [5.427346259545067]
  Chain-of-thought (CoT) has become central to scaling reasoning capabilities in large language models.<n>We show that instruction-tuned models often determine their answer before generating CoT.
  arXiv  Detail & Related papers  (2026-03-02T04:33:55Z)
• Decomposed Prompting Does Not Fix Knowledge Gaps, But Helps Models Say "I Don't Know" [47.930782177987446]
  Large language models often struggle to recognize their knowledge limits in closed-book question answering, leading to confident hallucinations.<n>We evaluate three task-equivalent prompting regimes: Direct, Assistive, and Incremental, across different model scales and multi-hop QA benchmarks.<n>Because factual knowledge is stable while hallucinations are agreement, cross-regime provides a precise signal of internal uncertainty.
  arXiv  Detail & Related papers  (2026-02-04T18:39:58Z)
• RAudit: A Blind Auditing Protocol for Large Language Model Reasoning [0.8594140167290097]
  Inference-time scaling can amplify reasoning pathologies: sycophancy, rung collapse, and premature certainty.<n>We present RAudit, a diagnostic protocol for auditing LLM reasoning without ground truth access.
  arXiv  Detail & Related papers  (2026-01-30T16:22:45Z)
• On the Paradoxical Interference between Instruction-Following and Task Solving [50.75960598434753]
  Instruction following aims to align Large Language Models (LLMs) with human intent by specifying explicit constraints on how tasks should be performed.<n>We reveal a counterintuitive phenomenon: instruction following can paradoxically interfere with LLMs' task-solving capability.<n>We propose a metric, SUSTAINSCORE, to quantify the interference of instruction following with task solving.
  arXiv  Detail & Related papers  (2026-01-29T17:48:56Z)
• Why and How LLMs Hallucinate: Connecting the Dots with Subsequence Associations [82.42811602081692]
  This paper introduces a subsequence association framework to systematically trace and understand hallucinations.<n>Key insight is hallucinations that arise when dominant hallucinatory associations outweigh faithful ones.<n>We propose a tracing algorithm that identifies causal subsequences by analyzing hallucination probabilities across randomized input contexts.
  arXiv  Detail & Related papers  (2025-04-17T06:34:45Z)
• Mitigating Knowledge Conflicts in Language Model-Driven Question Answering [15.29366851382021]
  Two fundamental knowledge sources play crucial roles in document-based question answering and document summarization systems.<n>Recent studies revealed a significant challenge: when there exists a misalignment between the model's inherent knowledge and the ground truth answers in training data, the system may exhibit problematic behaviors during inference.<n>Our investigation proposes a strategy to minimize hallucination by building explicit connection between source inputs and generated outputs.
  arXiv  Detail & Related papers  (2024-11-18T07:33:10Z)
• Context-Aware Reasoning On Parametric Knowledge for Inferring Causal Variables [49.31233968546582]
  We introduce a novel benchmark where the objective is to complete a partial causal graph.<n>We show the strong ability of LLMs to hypothesize the backdoor variables between a cause and its effect.<n>Unlike simple memorization of fixed associations, our task requires the LLM to reason according to the context of the entire graph.
  arXiv  Detail & Related papers  (2024-09-04T10:37:44Z)
• Nonparametric Identifiability of Causal Representations from Unknown Interventions [63.1354734978244]
  We study causal representation learning, the task of inferring latent causal variables and their causal relations from mixtures of the variables. Our goal is to identify both the ground truth latents and their causal graph up to a set of ambiguities which we show to be irresolvable from interventional data.
  arXiv  Detail & Related papers  (2023-06-01T10:51:58Z)
• Counterfactual Variable Control for Robust and Interpretable Question Answering [57.25261576239862]
  Deep neural network based question answering (QA) models are neither robust nor explainable in many cases. In this paper, we inspect such spurious "capability" of QA models using causal inference. We propose a novel approach called Counterfactual Variable Control (CVC) that explicitly mitigates any shortcut correlation.
  arXiv  Detail & Related papers  (2020-10-12T10:09:05Z)

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.