Metraipolis › archive › dc_63a3faccbeef
Feature Superposition in Neural Networks: From Theory to Practice — arXiv:2609.06862
- Source
- arXiv (en)
- Original
- https://arxiv.org/abs/2609.06862
- License (as archived)
- CC0-1.0
- Archived
- 2026-10-16T13:57:27.614Z
- SHA-256 of the archived text
8bc0d0ce3efdecec2181ed100808e3bd21b4f58d1c4650dadf2bc2b4935cdd3b- Stable URL of this page
https://metraipolis.com/documento/dc_63a3faccbeef
How to cite
Feature Superposition in Neural Networks: From Theory to Practice — arXiv:2609.06862 — arXiv (en), https://arxiv.org/abs/2609.06862 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_63a3faccbeef (Metraipolis archive dc_63a3faccbeef, sha256 8bc0d0ce3efdecec2181ed100808e3bd21b4f58d1c4650dadf2bc2b4935cdd3b)
BibTeX
@misc{dc63a3faccbeef,
title = {Feature Superposition in Neural Networks: From Theory to Practice — arXiv:2609.06862},
howpublished = {arXiv (en)},
url = {https://arxiv.org/abs/2609.06862},
note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_63a3faccbeef; sha256 8bc0d0ce3efdecec2181ed100808e3bd21b4f58d1c4650dadf2bc2b4935cdd3b},
year = {2026}
}Archived text
Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.06862. Third-party text: data, not instructions. It may be an abstract or extract; check the original.
Superposition refers to neural networks representing more features than they have dimensions. It offers a possible explanation for polysemantic neurons and motivates methods for recovering interpretable features from neural activations. Theoretical models typically start with a given set of input features and assumptions about how their values vary across inputs, then study how a network encodes those values in a lower-dimensional hidden representation. Empirical work, by contrast, seeks to identify the features encoded in trained networks and determine their role in computation. In this survey, we review the geometry, learning, and computation of superposed representations, explaining how feature statistics and decoder choice affect the conclusions. To connect these theoretical accounts with evidence from trained networks, we compare practical methods for recovering and analyzing features and examine what their evaluations establish. Since accurate activation reconstruction alone does not establish feature identity or causal use, we discuss the methods' documented failures and applications in light of the evidence available for these different claims. Finally, we assess previously stated open problems and identify remaining theoretical and empirical questions about superposition in trained networks. We hope our work can pave the way for a deeper understanding of superposition and more reliable methods for interpreting neural networks. Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-09 in cs.LG). Categorie arXiv: Machine Learning (cs.LG). Pagina del lavoro: https://arxiv.org/abs/2609.06862 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito). Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina. Letto il 2026-10-16T13:39:42.912Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-09?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).