{"slug": "learning-to-be-monitored-causal-bounds-and-intervention-aware-post-training-for", "title": "Learning to Be Monitored: Causal Bounds and Intervention-Aware Post-Training Recipes for Chain-of-Thought Monitorability", "summary": "A new arXiv paper (2610.31733v1) shows that minimal-criterion sensitivity (TPR_mc), the metric proposed by Guan et al. 2025 for intervention-based chain-of-thought monitoring evaluations, can overestimate how often a monitor detects target behaviors actually caused by an intervention, and establishes TPR_mc as an upper bound on that causal detection rate while deriving a complementary lower bound for a tighter evaluation framework. Using a training dataset of paired control-intervention tasks spanning multiple intervention types, the authors applied three intervention-specific post-training methods — off-policy self-distillation, reinforcement learning, and on-policy self-distillation — to separate variants of each base model, and report that all three generally improve CoT monitorability over the corresponding base models across model-monitor configurations. Code is available at github.com/xinyue-lxy/learning-to-be-monitored.", "body_md": "arXiv:2610.31733v1 Announce Type: new \nAbstract: Chain-of-thought (CoT) monitoring aims to detect potentially harmful behaviors in large language models (LLMs) from their CoTs. Intervention-based evaluations measure this property by asking whether a monitor can detect the role of an intervention from the CoT when the targeted behavior occurs, typically using minimal-criterion sensitivity ($\\mathsf{TPR_{mc}}$) proposed by~\\citep{guan2025monitoring}. However, a target behavior observed under intervention is not necessarily caused by it, and we show that this metric can overestimate how often a monitor detects target behaviors actually caused by the intervention. Specifically, we establish $\\mathsf{TPR_{mc}}$ as an upper bound on this causal detection rate and derive a complementary lower bound, which provides a tighter evaluation framework for CoT monitorability. With this framework, we study whether LLMs can be trained to make their CoTs easier to monitor. We first construct a training dataset of paired control--intervention tasks spanning multiple intervention types. Using the dataset, we design and apply three intervention-specific training methods based on off-policy self-distillation, reinforcement learning, and on-policy self-distillation to train separate variants of each base model. Across various model-monitor configurations, all three methods generally improve CoT monitorability over the corresponding base models under our extended evaluation framework. Our code is available at https://github.com/xinyue-lxy/learning-to-be-monitored.", "url": "https://wpnews.pro/news/learning-to-be-monitored-causal-bounds-and-intervention-aware-post-training-for", "canonical_source": "https://arxiv.org/abs/2610.31733", "published_at": "2026-10-27 04:00:00+00:00", "updated_at": "2026-10-27 04:13:08.259813+00:00", "lang": "en", "topics": ["ai-safety", "large-language-models", "ai-research", "machine-learning"], "entities": ["arXiv", "Guan et al. 2025", "TPR_mc", "github.com/xinyue-lxy/learning-to-be-monitored"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/learning-to-be-monitored-causal-bounds-and-intervention-aware-post-training-for", "markdown": "https://wpnews.pro/news/learning-to-be-monitored-causal-bounds-and-intervention-aware-post-training-for.md", "text": "https://wpnews.pro/news/learning-to-be-monitored-causal-bounds-and-intervention-aware-post-training-for.txt", "jsonld": "https://wpnews.pro/news/learning-to-be-monitored-causal-bounds-and-intervention-aware-post-training-for.jsonld"}}