Are Reasoning LLMs Robust to Interventions on Their Chain-of-Thought?

• URL: http://arxiv.org/abs/2602.07470v1
• Date: Sat, 07 Feb 2026 10:02:58 GMT
• Title: Are Reasoning LLMs Robust to Interventions on Their Chain-of-Thought?
• Authors: Alexander von Recum, Leander Girrbach, Zeynep Akata,
• Abstract summary: Reasoning LLMs generate step-by-step chains of thought before giving an answer.<n>How robust are these reasoning traces to disruptions that occur within them?<n>We introduce a controlled evaluation framework that perturbs a model's own CoT at fixed timesteps.
• Score: 79.86483056611105
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Reasoning LLMs (RLLMs) generate step-by-step chains of thought (CoTs) before giving an answer, which improves performance on complex tasks and makes reasoning more transparent. But how robust are these reasoning traces to disruptions that occur within them? To address this question, we introduce a controlled evaluation framework that perturbs a model's own CoT at fixed timesteps. We design seven interventions (benign, neutral, and adversarial) and apply them to multiple open-weight RLLMs across Math, Science, and Logic tasks. Our results show that RLLMs are generally robust, reliably recovering from diverse perturbations, with robustness improving with model size and degrading when interventions occur early. However, robustness is not style-invariant: paraphrasing suppresses doubt-like expressions and reduces performance, while other interventions trigger doubt and support recovery. Recovery also carries a cost: neutral and adversarial noise can inflate CoT length by more than 200%, whereas paraphrasing shortens traces but harms accuracy. These findings provide new evidence on how RLLMs maintain reasoning integrity, identify doubt as a central recovery mechanism, and highlight trade-offs between robustness and efficiency that future training methods should address.

Related papers

• Balancing Faithfulness and Performance in Reasoning via Multi-Listener Soft Execution [79.98699884805636]
  Reasoning Execution by Multiple Listeners (REMUL) is a multi-party reinforcement learning approach.<n>REMUL builds on the hypothesis that reasoning traces which other parties can follow will be more faithful.<n>Speakers are rewarded for producing reasoning that is clear to listeners.
  arXiv  Detail & Related papers  (2026-02-18T02:55:55Z)
• On Robustness and Chain-of-Thought Consistency of RL-Finetuned VLMs [15.301640007799735]
  We show that simple, controlled textual perturbations--misleading captions or incorrect chain-of-thought (CoT) traces--cause substantial drops in robustness and confidence.<n>To better understand these vulnerabilities, we analyze RL fine-tuning dynamics and uncover an accuracy-faithfulness trade-off.
  arXiv  Detail & Related papers  (2026-02-13T01:12:00Z)
• Stop Rewarding Hallucinated Steps: Faithfulness-Aware Step-Level Reinforcement Learning for Small Reasoning Models [59.6715047267181]
  Small reasoning models (SRMs) are prone to hallucinations, especially in intermediate reasoning steps.<n>Existing mitigation methods based on online reinforcement learning rely on outcome-based rewards or coarse-grained chain-of-thought evaluation.<n>We propose Faithfulness-Aware Step-Level Reinforcement Learning (FaithRL), introducing step-level supervision via explicit faithfulness rewards from a process reward model.
  arXiv  Detail & Related papers  (2026-02-05T17:15:12Z)
• Distractor Injection Attacks on Large Reasoning Models: Characterization and Defense [16.519353449118814]
  We analyze a critical vulnerability we term reasoning distraction, where LRMs are diverted from their primary objective by irrelevant yet complex tasks maliciously embedded in the prompt.<n>We show that even state-of-the-art LRMs are highly susceptible, with injected distractors reducing task accuracy by up to 60%.<n>We propose a training-based defense that combines Supervised Fine-Tuning (SFT) and Reinforcement Learning (RL) on synthetic adversarial data, improving robustness by over 50 points on challenging distractor attacks.
  arXiv  Detail & Related papers  (2025-10-17T23:16:34Z)
• From <Answer> to <Think>: Multidimensional Supervision of Reasoning Process for LLM Optimization [62.07990937720985]
  Dimension-level Reward Model (DRM) is a new supervision framework for Large Language Models.<n>DRM evaluates the quality of a reasoning process along three fundamental, complementary, and interpretable dimensions.<n> Experimental results show that DRM provides effective supervision signals, guides the optimization of LLMs and enhances their reasoning ability.
  arXiv  Detail & Related papers  (2025-10-13T14:29:15Z)
• Chain-of-Code Collapse: Reasoning Failures in LLMs via Adversarial Prompting in Code Generation [0.3495246564946556]
  Large Language Models (LLMs) have achieved remarkable success in tasks requiring complex reasoning.<n>Do these models truly reason, or do they merely exploit shallow statistical patterns?<n>We introduce Chain-of-Code Collapse, where we investigate the robustness of reasoning LLMs by introducing a suite of semantically faithful yet adversarially structured prompt perturbations.
  arXiv  Detail & Related papers  (2025-06-08T02:43:46Z)
• ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning [64.93140713419561]
  Large Reasoning Models (LRMs) perform strongly in complex reasoning tasks via Chain-of-Thought (CoT) prompting, but often suffer from verbose outputs.<n>Existing fine-tuning-based compression methods either operate post-hoc pruning, risking disruption to reasoning coherence, or rely on sampling-based selection.<n>We introduce ConCISE, a framework designed to generate concise reasoning chains, integrating Confidence Injection to boost reasoning confidence, and Early Stopping to terminate reasoning when confidence is sufficient.
  arXiv  Detail & Related papers  (2025-05-08T01:40:40Z)
• Dissociation of Faithful and Unfaithful Reasoning in LLMs [2.4893095725361922]
  We investigate how large language models (LLMs) recover from errors in Chain of Thought. We find evidence for unfaithfulness in Chain of Thought, which occurs when models arrive at the correct answer despite invalid reasoning text.
  arXiv  Detail & Related papers  (2024-05-23T22:38:58Z)
• Seeing is not Believing: Robust Reinforcement Learning against Spurious Correlation [57.351098530477124]
  We consider one critical type of robustness against spurious correlation, where different portions of the state do not have correlations induced by unobserved confounders. A model that learns such useless or even harmful correlation could catastrophically fail when the confounder in the test case deviates from the training one. Existing robust algorithms that assume simple and unstructured uncertainty sets are therefore inadequate to address this challenge.
  arXiv  Detail & Related papers  (2023-07-15T23:53:37Z)

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.