Methods · Comparison with Alternative Feature Extraction Methods
Supervised Feature Extraction Methods (DAS, SMDS, probes)
Survey sections: §6.9.4 Supervised Feature Extraction Methods
When the concepts of interest are known in advance, supervised methods can directly optimise representations aligned with those targets.
Intended applications
When labelled data is available, supervised approaches often outperform SAEs on concept detection and steering tasks (Huang et al., 2024; Wu et al., 2025b; Chaudhary & Geiger, 2024). They also avoid choosing dictionary sizes or sparsity penalties, since the concept set determines complexity. However, these methods require labelled data for each concept and cannot discover previously unknown features. Therefore, supervised methods suit settings where predefined, crisp concepts matter, whereas SAEs remain preferable for exploratory analysis.
Consider alternatives for targeted analysis: When specific concepts are the focus, supervised methods (e.g., DAS, probes, discussed in Section 6.9.4) may be more reliable because SAEs discover features in an unsupervised and exploratory manner.
Related methods
Key papers
Cited in §6.9.4.
- Atticus Geiger, Zhengxuan Wu, Christopher Potts, Thomas Icard, and Noah Goodman. Finding alignments between interpretable causal variables and distributed neural representations. In Francesco Locatello and Vanessa Didelez (eds.), Proceedings of the Third Conference on Causal Learning and Reasoning, volume 236 of Proceedings of Machine Learning Research, pp. 160–187. PMLR, 01–03 Apr 2024. URL https://proceedings.mlr.press/v236/geiger24a.html.
- Federico Tiblias, Irina Bigoulaeva, Jingcheng Niu, Simone Balloccu, and Iryna Gurevych. Hypothesis-driven feature manifold analysis in LLMs via supervised multi-dimensional scaling. Transactions on Machine Learning Research, 2026. ISSN 2835-8856. URL https://openreview.net/forum?id=vCKZ40YYPr.
- Samuel Marks and Max Tegmark. The geometry of truth: Emergent linear structure in large language model representations of true/false datasets. In First Conference on Language Modeling, 2024. URL https://openreview.net/forum?id=aajyHYjjsk.
- Zhengxuan Wu, Aryaman Arora, Atticus Geiger, Zheng Wang, Jing Huang, Dan Jurafsky, Christopher D Manning, and Christopher Potts. Axbench: Steering LLMs? even simple baselines outperform sparse autoencoders. In Forty-second International Conference on Machine Learning, 2025b. URL https://openreview.net/forum?id=K2CckZjNy0.
- Andy Arditi, Oscar Balcells Obeso, Aaquib Syed, Daniel Paleka, Nina Rimsky, Wes Gurnee, and Neel Nanda. Refusal in language models is mediated by a single direction. In The Thirty-eighth Annual Conference on Neural Information Processing Systems, 2024.
- PreprintMonte MacDiarmid, Timothy Maxwell, Nicholas Schiefer, Jesse Mu, Jared Kaplan, David Duvenaud, Sam Bowman, Alex Tamkin, Ethan Perez, Mrinank Sharma, Carson Denison, and Evan Hubinger. Simple probes can catch sleeper agents. https://www.anthropic.com/news/probes-catch-sleeper-agents, 2024.
- Jing Huang, Zhengxuan Wu, Christopher Potts, Mor Geva, and Atticus Geiger. RAVEL: Evaluating interpretability methods on disentangling language model representations. In Lun-Wei Ku, Andre Martins, and Vivek Srikumar (eds.), Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pp. 8669–8687, Bangkok, Thailand, August 2024. Association for Computational Linguistics. . URL https://aclanthology.org/2024.acl-long.470/.
- PreprintMaheep Chaudhary and Atticus Geiger. Evaluating open-source sparse autoencoders on disentangling factual knowledge in gpt-2 small, 2024. URL https://arxiv.org/abs/2409.04478.
Also discussed in: §6.4 Analysis Goal · §6.5.6 Practical Recommendations · §9.3 Steering as a Causal Test for Feature Interpretability
Other methods in Comparison with Alternative Feature Extraction Methods
Non-Negative Matrix Factorisation (NMF) · Principal Component Analysis (PCA) · Independent Component Analysis (ICA) · Dictionary Learning Baselines (K-SVD)