对比多种大模型控制方法，发现提示词效果最好，稀疏自编码器不占优。
AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders
- 构建AxBench基准，系统评估多种语言模型控制技术。
- 提示词在输出控制上超越所有方法，微调次之，稀疏自编码器表现最差。
- 提出新方法ReFT-r1，在可控性与可解释性间取得平衡，适合安全应用。
精细控制大语言模型输出对安全与可靠性至关重要。尽管提示词和微调广泛应用，但解释性研究还提出了多种基于表示的技术，如稀疏自编码器（SAEs）、线性人工断层成像、监督控制向量、线性探测器和表示微调。目前尚无直接比较这些方法的基准。为此，我们引入AxBench，一个大规模的控制与概念检测基准，并在Gemma-2-2B和9B上进行实验。结果显示，在控制任务中，提示词优于所有现有方法，其次是微调；在概念检测任务中，基于表示的方法如均值差（Difference-in-Means）表现最佳。在两项任务中，SAEs均不具备竞争力。我们提出一种新型弱监督表示方法（Rank-1表示微调；ReFT-r1），在两项任务中均表现良好，同时具备提示词所缺乏的可解释性优势。随AxBench一同发布，我们训练并公开了ReFT-r1和DiffMean的SAE规模特征字典。
原文摘要 · Abstract (English)
Fine-grained steering of language model outputs is essential for safety and reliability. Prompting and finetuning are widely used to achieve these goals, but interpretability researchers have proposed a variety of representation-based techniques as well, including sparse autoencoders (SAEs), linear artificial tomography, supervised steering vectors, linear probes, and representation finetuning. At present, there is no benchmark for making direct comparisons between these proposals. Therefore, we introduce AxBench, a large-scale benchmark for steering and concept detection, and report experiments on Gemma-2-2B and 9B. For steering, we find that prompting outperforms all existing methods, followed by finetuning. For concept detection, representation-based methods such as difference-in-means, perform the best. On both evaluations, SAEs are not competitive. We introduce a novel weakly-supervised representational method (Rank-1 Representation Finetuning; ReFT-r1), which is competitive on both tasks while providing the interpretability advantages that prompting lacks. Along with AxBench, we train and publicly release SAE-scale feature dictionaries for ReFT-r1 and DiffMean.
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。