外观
论文arXiv 2610.1749516 天前
语言模型中的本体感
原文链接：A Sense of Proprioception in Language Models（arXiv 2610.17495 · 发布日期：2026-10-12）
A Sense of Proprioception in Language Models
本文探讨语言模型感知内部状态的问题，通过SFT和on-policy训练赋予其数字本体感，并在多类行为评估中验证影响。
核心要点
- 动机是语言模型训练时无法观察并准确报告自身内部状态。
- 方法将模型内部信息构造成可验证信号，并用SFT和on-policy训练学习。
- 结果显示现代模型可学习路由决策、token熵、决策深度等本体感子类型。
- 进一步评估本体感访问对情境意识、可读性和附带现象性的影响。
验证范围与局限：验证范围为现代语言模型上的训练与行为评估，摘要未说明具体模型或任务。
英文摘要 (Abstract)
In humans, proprioception corresponds to the sense of self-movement or position, e.g. understanding the position of one's arm. This sense is often less well-understood and in popular understanding sometimes conflated with introspection. But, notably, proprioception is a separate sense, with dedicated receptors and verifiable outcomes. On the other hand though, language models cannot observe their own internal states during training (and for example answer the question which experts activated in a certain layer of an MoE) and so cannot accurately report on them. Yet, such information is, by virtue of being internal to the model, both accessible, and readily constructed into verifiable signals. We argue that this provides the basis for a digital analog of proprioception. As such, we train models through SFT and on-policy training to equip them with this ability. We show that modern models can be trained to understand a range of subtypes of proprioception, such as routing decision awareness, token entropy, decision depth or logit lens information, without affecting model capabilities. We then run these models through a broad evaluation of situational awareness, legibility, and epiphenomenality, to evaluate in what ways access to proprioception changes model behavior.
作者：Jonas Geiping
来源
本页事实对应的公开页面，以原文为准。