AI模型内部有个“隐形时钟”在倒计时
扩散语言模型（DLM）是一种新型AI，它生成文本时不像传统模型那样一步步“说”出来，而是从一团噪声中逐渐“洗”出答案。研究者发现，这类模型内部竟然藏着一个“隐形时钟”——它虽然没有被明确告知当前处于第几步，但模型自己默默记录着去噪进度，并且这个信号可以被提取出来。更神奇的是，通过微调这个“时钟”信号，可以控制模型输出的确定性和随机性。这就像你发现一个黑箱里有个隐藏的秒表，而且你还能拨动它来改变结果。虽然你明天用不上，但它揭示了AI内部一种意想不到的自我组织能力。
📄 原文摘要(英文)
Diffusion Language Models (DLMs) have recently emerged as a promising alternative to autoregressive models. Unlike standard diffusion-based approaches, DLMs are not explicitly conditioned on a timestep, raising a natural question: do these models internally represent denoising progress, and how is such information used downstream? In this work, we show that DLMs do in fact encode a latent representation related to the diffusion timestep within their residual streams. We find that this signal can be reliably extracted using probes across layers, indicating that denoising progress is decodable from internal activations. We further demonstrate that steering the model along a low-dimensional subspace associated with the inferred timestep allows us to systematically modulate its notion of denoising progress, leading to predictable changes in model confidence and entropy. Finally, we analyse the geometry of the identified representation, showing that it exhibits structured and interpretable properties in activation space, and shedding light on how such a signal is processed by these models.