paper-with-me

Papers

When Chain of Thought is Necessary, Language Models Struggle to Evade Monitors

2025-07-07 · Scott Emmons, Erik Jenner, David K. Elson, Rif A. Saurous, Senthooran Rajamanoharan, Heng Chen, Irhum Shafkat, Rohin Shah arxiv

While chain-of-thought (CoT) monitoring is an appealing AI safety defense, recent work on "unfaithfulness" has cast doubt on its reliability. These findings highlight an important failure mode, particularly when CoT acts as a post-hoc rationalization in applications like auditing for bias. However, for the distinct problem of runtime monitoring to prevent severe harm, we argue the key property is not faithfulness but monitorability. To this end, we introduce a conceptual framework distinguishing CoT-as-rationalization from CoT-as-computation. We expect that certain classes of severe harm will require complex, multi-step reasoning that necessitates CoT-as-computation. Replicating the experimental setups of prior work, we increase the difficulty of the bad behavior to enforce this necessity condition; this forces the model to expose its reasoning, making it monitorable. We then present methodology guidelines to stress-test CoT monitoring against deliberate evasion. Applying these guidelines, we find that models can learn to obscure their intentions, but only when given significant help, such as detailed human-written strategies or iterative optimization against the monitor. We conclude that, while not infallible, CoT monitoring offers a substantial layer of defense that requires active protection and continued stress-testing.

📄 PDF Abstract BibTeX arXiv:2507.05246

Code (0)

등록된 구현이 없습니다.

Similar Papers 제목 키워드 기반

RIPPLECOT: Amplifying Ripple Effect of Knowledge Editing in Language Models via Chain-of-Thought In-Context Learning

2024-10-04 · Zihao Zhao, Yuchen Yang, Yijiang Li, Yinzhi Cao

The ripple effect poses a significant challenge in knowledge editing for large language models. Namely, when a single fact is edited, the model struggles to accurately update the related facts in a sequence, which is eva…

In-Context Learningknowledge editing

Chain-of-Thought Augmentation with Logit Contrast for Enhanced Reasoning in Language Models

2024-07-04 · Jay Shim, Grant Kruttschnitt, Alyssa Ma, Daniel Kim 외

Rapidly increasing model scales coupled with steering methods such as chain-of-thought prompting have led to drastic improvements in language model reasoning. At the same time, models struggle with compositional generali…

Language ModelingLanguage Modelling

Interleaving Retrieval with Chain-of-Thought Reasoning for Knowledge-Intensive Multi-Step Questions

2022-12-20 · Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, Ashish Sabharwal

Prompting-based large language models (LLMs) are surprisingly powerful at generating natural language reasoning steps or Chains-of-Thoughts (CoT) for multi-step question answering (QA). They struggle, however, when the n…

HallucinationQuestion AnsweringRetrieval

Tree of Thoughts: Deliberate Problem Solving with Large Language Models

2023-05-17 · NeurIPS 2023 11 · Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran 외

Language models are increasingly being deployed for general problem solving across a wide range of tasks, but are still confined to token-level, left-to-right decision-making processes during inference. This means they c…

Arithmetic ReasoningDecision MakingLanguage ModellingQuestion Answering

Can Confidence Estimates Decide When Chain-of-Thought Is Necessary for LLMs?

2025-10-23 · Samuel Lewis-Lim, Xingwei Tan, Zhixue Zhao, Nikolaos Aletras arxiv

Chain-of-thought (CoT) prompting is a common technique for improving the reasoning abilities of large language models (LLMs). However, extended reasoning is often unnecessary and substantially increases token usage. As s…