用稀疏自编码器清除语言模型控制向量中的噪声，提升控制精度。
Denoising Concept Vectors with Sparse Autoencoders for Improved Language Model Steering
- 通过选择性保留最区分度高的编码潜变量，重构隐藏表示。
- 在6个挑战性概念上，控制成功率提升4-16%。
- 适合需要精准可控的生成任务研究者使用。
线性概念向量能有效操控大语言模型，但现有方法在多样化数据集中受噪声特征影响，降低控制鲁棒性。本文提出稀疏自编码器去噪概念向量（SDCV），通过选择性保留最具区分性的SAE潜变量，并重构隐藏表示来分离信号与噪声。关键思路是放大表现最佳的top-k潜变量激活值，以明确区分正负样本。应用于线性探测和均值差异法时，SDCV在六个挑战性概念上均稳定提升控制成功率4-16%，同时保持主题相关性。
原文摘要 · Abstract (English)
Linear concept vectors effectively steer LLMs, but existing methods suffer from noisy features in diverse datasets that undermine steering robustness. We propose Sparse Autoencoder-Denoised Concept Vectors (SDCV), which selectively keep the most discriminative SAE latents while reconstructing hidden representations. Our key insight is that concept-relevant signals can be explicitly separated from dataset noise by scaling up activations of top-k latents that best differentiate positive and negative samples. Applied to linear probing and difference-in-mean, SDCV consistently improves steering success rates by 4-16\% across six challenging concepts, while maintaining topic relevance.
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。