arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2411.07071cs.LGcs.AIcs.CL
收藏

LLMs中归纳的普遍响应与涌现

Universal Response and Emergence of Induction in LLMs

University of Hamburg(汉堡大学)

Niclas Luick

AI总结:

通过探测LLMs对残差流弱单token扰动的普遍且尺度不变的响应,研究归纳行为在中间层的涌现,并识别构成该行为的模型部分,为大规模电路分析提供基准。

摘要:

虽然归纳被认为是LLMs上下文学习的关键机制,但超越玩具模型理解其精确的电路分解仍然难以捉摸。在这里,我们通过探测LLMs对残差流的弱单token扰动响应,研究LLMs内归纳行为的涌现。我们发现LLMs表现出一种稳健的普遍机制,其响应在扰动强度变化时保持尺度不变,从而使我们能够量化整个模型中token相关性的积累。通过应用我们的方法,我们在Gemma-2-2B、Llama-3.2-3B和GPT-2-XL的残差流中观察到归纳行为的特征。在所有模型中,我们发现这些归纳特征在中间层逐渐涌现,并识别出构成该行为的相关模型部分。我们的结果为LLMs内组件的集体相互作用提供了见解,并可作为大规模电路分析的基准。

英文摘要:

While induction is considered a key mechanism for in-context learning in LLMs, understanding its precise circuit decomposition beyond toy models remains elusive. Here, we study the emergence of induction behavior within LLMs by probing their response to weak single-token perturbations of the residual stream. We find that LLMs exhibit a robust, universal regime in which their response remains scale-invariant under changes in perturbation strength, thereby allowing us to quantify the build-up of token correlations throughout the model. By applying our method, we observe signatures of induction behavior within the residual stream of Gemma-2-2B, Llama-3.2-3B, and GPT-2-XL. Across all models, we find that these induction signatures gradually emerge within intermediate layers and identify the relevant model sections composing this behavior. Our results provide insights into the collective interplay of components within LLMs and serve as a benchmark for large-scale circuit analysis.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑