arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2510.01246cs.CL
收藏

语言模型调控中稀疏自编码器与激活差异的对比分析

A Comparative Analysis of Sparse Autoencoder and Activation Difference in Language Model Steering

Jiaqing Xie

AI总结:

针对语言模型调控中top-k SAE隐变量含非语义特征、恒定调控易产生退化输出的问题,提出top-1单隐变量聚焦与逐token衰减调控策略,实验证明其在数学推理基准上优于平均激活差异方法,IF-Eval表现相当。

摘要:

稀疏自编码器(Sparse Autoencoders, SAEs)近期已成为语言模型调控的有力工具。已有研究探索了使用top-k个SAE隐变量进行调控,但我们观察到,top-k隐变量中的许多维度捕捉的是标点等非语义特征,而非指令这类语义属性。为解决该问题,我们提出聚焦于单个最相关的SAE隐变量(top-1),以消除冗余特征。我们进一步发现恒定SAE调控存在局限性,其常会产生重复单字之类的退化输出。为缓解该问题,我们提出一种逐token衰减的调控策略,使其能与平均激活差异基线开展更可靠的对比。实验结果表明,调控与推理相关的SAE隐变量能可靠地触发逐步数学推理并提升推理质量,其功能效果类似于附加引导token。我们的结果显示,SAEs在数学推理基准上的表现优于平均激活差异方法,在IF-Eval上的表现与之相当。

英文摘要:

Sparse autoencoders (SAEs) have recently emerged as a powerful tool for language model steering. Prior work has explored top-k SAE latents for steering, but we observe that many dimensions among the top-k latents capture non-semantic features such as punctuation rather than semantic attributes like instructions. To address this, we propose focusing on a single, most relevant SAE latent (top-1), eliminating redundant features. We further identify a limitation in constant SAE steering, which often produces degenerate outputs such as repetitive single words. To mitigate this, we introduce a token-wise decaying steering strategy, enabling more faithful comparisons with mean activation difference baselines. Empirically, we show that steering an SAE latent associated with reasoning reliably elicits step-by-step mathematical reasoning and enhances inference quality, functionally resembling the effect of appending a guiding token. Our results demonstrate that SAEs outperform mean activation difference methods on mathematical reasoning benchmarks and match their performance on IF-Eval.

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。

↑