On the Chain-of-Thought Monitorability of Looped Language Models

2026-10-01

AI Summary

This study addresses the unclear impact of Loop Language Models (LoopLMs) on the monitorability of Chain-of-Thought (CoT) reasoning. We present the first systematic evaluation of this problem using the MonitorBench benchmark under both standard and stress-testing settings, comparing the reasoning performance across varying loop depths against non-looping baselines alongside qualitative diagnostic analysis. Our findings reveal that while deep looping degrades CoT monitorability in certain logical and scientific tasks, the looping architecture itself does not inherently compromise overall monitoring performance. By elucidating the differential effects of looping mechanisms on model transparency, this work provides critical empirical evidence to guide the design of safe and controllable looping architectures.

Abstract

Chain-of-thought (CoT) monitoring provides a promising approach for detecting undesirable model behavior. Looped language models (LoopLMs) repeatedly apply shared transformer layers, increasing effective computational depth and enabling additional latent computation without increasing model size. However, the effect of looped architectures on CoT monitorability remains largely unexplored. In this work, we provide the first systematic evaluation of CoT monitorability in LoopLMs. We study two complementary settings: (1) varying the loop depth within the same LoopLM family to isolate the effect of additional recurrent computation, and (2) comparing LoopLMs with non-looped language models matched by parameter size, transformer-layer count, or effective depth to study whether LoopLMs are less monitorable. Across eight tasks from MonitorBench and both standard and stress-test settings, we observe task-dependent reductions in CoT monitorability under stress tests on specific Logic/Science/Engineering \texttt{Cue Answer} tasks, while other tasks exhibit weaker or qualitatively different trends. Our diagnosis suggests that these declines are not fully explained by task difficulty, verification pass rate, or generated token length; qualitative examples further suggest changes in how deeper-loop models explicitly use or attribute provided cues. Our cross-model comparison finds no evidence that LoopLMs are systematically less monitorable than non-looped language models matched on size or depth. Overall, our results suggest that deeper loop depth can reduce CoT monitorability in some tasks under stress tests, but looped transformer architecture alone does not necessarily imply lower monitorability.

Problem

• Chain-of-Thought Monitorability
• Looped Language Models
• Recurrent Computation
• AI Safety

Innovation

• Chain-of-Thought Monitorability
• Looped Language Models
• Recurrent Computation
• Stress Testing
• Cross-model Comparison