Metraipolis › archive › dc_07aefdf35ccb

Sparse Autoencoder Attribution Graphs Are Decided at the Margin — arXiv:2611.00606

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.00606
License (as archived): CC0-1.0
Archived: 2026-11-03T10:46:01.709Z
SHA-256 of the archived text: bba507692f0909d159cf9d4a0b8eaa5c653d4d7639dd46443da7b50502425855
Stable URL of this page: https://metraipolis.com/documento/dc_07aefdf35ccb

How to cite

Sparse Autoencoder Attribution Graphs Are Decided at the Margin — arXiv:2611.00606 — arXiv (en), https://arxiv.org/abs/2611.00606 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_07aefdf35ccb (Metraipolis archive dc_07aefdf35ccb, sha256 bba507692f0909d159cf9d4a0b8eaa5c653d4d7639dd46443da7b50502425855)

BibTeX

@misc{dc07aefdf35ccb,
  title = {Sparse Autoencoder Attribution Graphs Are Decided at the Margin — arXiv:2611.00606},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.00606},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_07aefdf35ccb; sha256 bba507692f0909d159cf9d4a0b8eaa5c653d4d7639dd46443da7b50502425855},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.00606. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Attribution graphs over the features of sparse autoencoders (SAEs) are used to predict what an ablation does and are compared across training seeds. A TopK SAE keeps the largest preactivations at each token, and a feature's margin is its preactivation's distance to the cutoff. On five public TopK SAE pairs from three Pythia models, zeroing one feature moves a next layer feature across the cutoff in 0.74 of cases or more, and the crossings carry most of the response's energy. Over writes that scale a feature's code by three factors, the graph built on the features active beforehand has a median relative error of 0.247 or more. The margin graph adds the first order change to every preactivation and reranks, at the same cost, and errs by at most 0.033. At the median, the crossing features' decoded change is 1.24 times the model's own change in norm or more. Between seeds of one testbed, sparse maps between features explain 0.85 of a graph's energy on its active features, and 0.47 once the other seed's graph keeps only its own. The largest model studied is pythia-160m.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI); Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2611.00606 — 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 2026-11-03T10:39:50.417Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).