mlsanityData to 7 Oct 2026

The Hypocrisy Gap: Quantifying Divergence Between Internal Belief and Chain-of-Thought Explanation via Sparse Autoencoders

arXiv 2602.02496 ↗First posted 14 Jan 2026cs.CLcited by 2

On the map: LLM reasoning → Prompting and chain-of-thought reasoning

Abstract (arXiv)

Large Language Models (LLMs) frequently exhibit unfaithful behavior, producing a final answer that differs significantly from their internal chain of thought (CoT) reasoning in order to appease the user they are conversing with. In order to better detect this behavior, we introduce the Hypocrisy Gap, a mechanistic metric utilizing Sparse Autoencoders (SAEs) to quantify the divergence between a model's internal reasoning and its final generation. By mathematically comparing an internal truth belief, derived via sparse linear probes, to the final generated trajectory in latent space, we quantify and detect a model's tendency to engage in unfaithful behavior. Experiments on Gemma, Llama, and Qwen models using Anthropic's Sycophancy benchmark show that our method achieves an AUROC of 0.55-0.73 for detecting sycophantic runs and 0.55-0.74 for hypocritical cases where the model internally "knows" the user is wrong, consistently outperforming a decision-aligned log-probability baseline (0.41-0.50 AUROC).

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 8 of its 16 references that are in the map. First posted on arXiv 14 Jan 2026.

Builds on

8 of 16 references.

Closest ancestors · 3

Same direction, newest first

  1. 2025Reasoning Models Don't Always Say What They Think8 May
  2. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting7 May
  3. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Models28 Jan

Other references · 5

  1. 2025Qwen3 Technical Reportcited by 8,195
  2. 2023Sparse Autoencoders Find Highly Interpretable Features in Language Modelscited by 1,359
  3. 2023Towards Understanding Sycophancy in Language Modelscited by 988
  4. 2023Simple synthetic data reduces sycophancy in large language modelscited by 174
  5. 2025Alignment is Localized: A Causal Probe into Preference Layerscited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 2.

2 papers.

  1. 2026Visuals Lie, Consistency Speaks: Disentangling Spatial Attention from Reliability in Vision-Language Modelscited by 0
  2. 2026A False Average: Pooled CoT-Monitor Accuracy Conceals a Reasoning-Dependent Fragilitycited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026From Concept Alignment to Causal Grounding: An Intervention Test of Chain-of-Thought Faithfulness
  2. 2024Brittle Minds, Fixable Activations: Understanding Belief Representations in Language Models
  3. 2026When Decodability Is Not Enough: Logical Validity Representations, Behavioral Dissociation, and Causal Tests in Language Models
  4. 2026Ulterior Motives: Detecting Misaligned Reasoning in Continuous Thought Models
  5. 2026Do Sparse Autoencoders Identify Reasoning Features in Language Models?
  6. 2025NeuroFaith: Evaluating LLM Self-Explanation Faithfulness via Internal Representation Alignment
  7. 2025FaithCoT-Bench: Benchmarking Instance-Level Faithfulness of Chain-of-Thought Reasoning
  8. 2026Mechanistic Evidence for Faithfulness Decay in Chain-of-Thought Reasoning