mlsanityData to 14 Oct 2026

Selective Disclosure of Hidden Directives in Reasoning Models: Behavioral Asymmetry and Steering

arXiv 2608.29070 ↗First posted 29 Aug 2026cs.LGcited by 0

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

Abstract (arXiv)

Chain-of-thought (CoT) reasoning traces are increasingly proposed as a mechanism for AI oversight: a monitor inspecting a model's reasoning can, in principle, detect misbehavior invisible from outputs alone. This assumes CoT surfaces what a model is instructed to do regardless of the instructions given. We test this assumption along two axes. First, we introduce the Instruction-Compliance Gap (ICG): the difference in probability that a model's CoT explicitly references a hidden system prompt directive when that directive is malign versus benign. Across 100 task pairs and 8 frontier reasoning models from 5 families, we find consistent asymmetric disclosure, a higher probability of leaking malign hidden instructions than benign ones, in Qwen3-14B (Wilcoxon p=0.0001, +13.9pp), Qwen3-32B (p=0.0011, +13.0pp), Qwen3-235B (p=0.035, +5.8pp), and similar results with MiniMax-M2.5 and DeepSeek-R1. The detector has 100% precision against two independent blinded labelling passes, and an LLM monitor reading only the reasoning trace reproduces the asymmetry in all 8 models against directive-free controls, identifying the specific directive in 82% of malign traces which the detector classifies as clean. Second, steering vectors extracted in MiniMax-M2.5 via Contrastive Activation Addition causally induce hiding from bare prompts and suppress it from prompts that would otherwise produce it, replicating in Qwen3-14B under a pre-registered design. Benign and malign-derived hiding vectors are highly similar (cosine 0.804 in MiniMax-M2.5; 0.970 in Qwen3-14B), implying that in these models the disclosure asymmetry arises from differential activation of a shared hiding direction rather than separate mechanisms.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Prompting and chain-of-thought reasoning, one of 17 directions in LLM reasoning. Builds on 13 of its 15 references that are in the map. First posted on arXiv 29 Aug 2026.

Builds on

13 of 15 references.

Closest ancestors · 7

Same direction, newest first

  1. 2025Can Reasoning Models Obfuscate Reasoning? Stress-Testing Chain-of-Thought Monitorability21 Oct
  2. 2025Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety15 Jul
  3. 2025When Thinking LLMs Lie: Unveiling the Strategic Deception in Representations of Reasoning Models5 Jun
  4. 2025Monitoring Reasoning Models for Misbehavior and the Risks of Promoting Obfuscation14 Mar
  5. 2023Measuring Faithfulness in Chain-of-Thought Reasoning17 Jul
  6. 2023Language Models Don't Always Say What They Think: Unfaithful Explanations in Chain-of-Thought Prompting7 May
  7. 2022Chain-of-Thought Prompting Elicits Reasoning in Large Language Models28 Jan

Other references · 6

  1. 2023Representation Engineering: A Top-Down Approach to AI Transparencycited by 1,336
  2. 2023Steering Llama 2 via Contrastive Activation Additioncited by 1,085
  3. 2022Discovering Latent Knowledge in Language Models Without Supervisioncited by 767
  4. 2024Frontier Models are Capable of In-context Schemingcited by 287
  5. 2023Large Language Models can Strategically Deceive their Users when Put Under Pressurecited by 146
  6. 2024SelfIE: Self-Interpretation of Large Language Model Embeddingscited by 67

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Reasoning Models Don't Always Say What They Think
  2. 2026Evading Chain-of-Thought Monitoring Through Model Poisoning
  3. 2026Bypassing the Rationale: Causal Auditing of Implicit Reasoning in Language Models
  4. 2025When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors
  5. 2026Detecting Hidden Chain-of-Thought in Large Language Models with Linguistic, Behavioral, and Mechanistic Indicators
  6. 2026Hidden Reasoning Must Leak, but Need Not Be Readable: Fundamental Opportunities and Limits for Chain-of-Thought Monitoring
  7. 2025Investigating CoT Monitorability in Large Reasoning Models
  8. 2026Noticing the Watcher: LLM Agents Can Infer CoT Monitoring from Blocking Feedback