轻量级监控架构可有效减少大模型推理错误，尤其适合复杂任务。
The cognitive companion: a lightweight parallel monitoring architecture for detecting and recovering from reasoning degradation in LLM agents
- 采用并行双架构监控，支持零开销探针与大模型助手
- 对易循环任务减少52%-62%重复，探针准确率达84%
- 适用于复杂开放任务，小模型效果不明显
大型语言模型在多步任务中常出现推理退化、循环、漂移、卡死等问题，难任务下发生率高达30%。现有方案或设硬性步骤限制（突兀），或用大模型做裁判（每步10-15%开销）。本文提出认知伙伴（Cognitive Companion）——一种并行监控架构，包含基于大模型的同伴与新型零开销探针同伴。实验基于Gemma 4 E4B进行三批可行性研究，并对Qwen 2.5 1.5B和Llama 3.2 1B小模型进行探索性分析。基于大模型的同伴在循环任务中降低52%-62%重复，开销约11%；探针同伴在第28层隐藏状态上训练，平均效应量+0.471，推理开销为零；其最佳结果在小型标注数据集上达到交叉验证AUROC 0.840。关键发现：同伴帮助显著依赖任务类型——在循环与开放任务中表现优异，而在结构化任务中无益甚至有害。小模型实验还暗示规模边界：1B-1.5B模型即使触发干预也未提升质量代理指标。论文应视为可行性研究，而非定论。结果表明子词级监控可能有效，任务类型敏感是设计约束，选择性激活是未来重要方向。
原文摘要 · Abstract (English)
Large language model (LLM) agents on multi-step tasks suffer reasoning degradation, looping, drift, stuck states, at rates up to 30% on hard tasks. Current solutions include hard step limits (abrupt) or LLM-as-judge monitoring (10-15% overhead per step). This paper introduces the Cognitive Companion, a parallel monitoring architecture with two implementations: an LLM-based Companion and a novel zero-overhead Probe-based Companion. We report a three-batch feasibility study centered on Gemma 4 E4B, with an additional exploratory small-model analysis on Qwen 2.5 1.5B and Llama 3.2 1B. In our experiments, the LLM-based Companion reduced repetition on loop-prone tasks by 52-62% with approximately 11% overhead. The Probe-based Companion, trained on hidden states from layer 28, showed a mean effect size of +0.471 at zero measured inference overhead; its strongest probe result achieved cross-validated AUROC 0.840 on a small proxy-labeled dataset. A key empirical finding is that companion benefit appears task-type dependent: companions are most helpful on loop-prone and open-ended tasks, while effects are neutral or negative on more structured tasks. Our small-model experiments also suggest a possible scale boundary: companions did not improve the measured quality proxy on 1B-1.5B models, even when interventions fired. Overall, the paper should be read as a feasibility study rather than a definitive validation. The results provide encouraging evidence that sub-token monitoring may be useful, identify task-type sensitivity as a practical design constraint, and motivate selective companion activation as a promising direction for future work.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。