AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2025
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866913717597241344 |
|---|---|
| author | Wu, Zhengxuan Arora, Aryaman Geiger, Atticus Wang, Zheng Huang, Jing Jurafsky, Dan Manning, Christopher D. Potts, Christopher |
| author_facet | Wu, Zhengxuan Arora, Aryaman Geiger, Atticus Wang, Zheng Huang, Jing Jurafsky, Dan Manning, Christopher D. Potts, Christopher |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2501_17148 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders Wu, Zhengxuan Arora, Aryaman Geiger, Atticus Wang, Zheng Huang, Jing Jurafsky, Dan Manning, Christopher D. Potts, Christopher Computation and Language Artificial Intelligence Machine Learning I.2.7 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. |
| title | AxBench: Steering LLMs? Even Simple Baselines Outperform Sparse Autoencoders |
| topic | Computation and Language Artificial Intelligence Machine Learning I.2.7 |
| url | https://arxiv.org/abs/2501.17148 |