Metraipolis › archive › dc_5fdfd1b7ca72

When Is an SAE Feature Interpretable? A Validation Ladder for EEG Foundation Models — arXiv:2609.34091

Source: arXiv (en)
Original: https://arxiv.org/abs/2609.34091
License (as archived): CC0-1.0
Archived: 2026-10-07T22:37:59.723Z
SHA-256 of the archived text: 9197d4066144a14f41f5fdd58ea9219f14c3c9977957d2f8cbbc204695c5beb0
Stable URL of this page: https://metraipolis.com/documento/dc_5fdfd1b7ca72

How to cite

When Is an SAE Feature Interpretable? A Validation Ladder for EEG Foundation Models — arXiv:2609.34091 — arXiv (en), https://arxiv.org/abs/2609.34091 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_5fdfd1b7ca72 (Metraipolis archive dc_5fdfd1b7ca72, sha256 9197d4066144a14f41f5fdd58ea9219f14c3c9977957d2f8cbbc204695c5beb0)

BibTeX

@misc{dc5fdfd1b7ca72,
  title = {When Is an SAE Feature Interpretable? A Validation Ladder for EEG Foundation Models — arXiv:2609.34091},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34091},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_5fdfd1b7ca72; sha256 9197d4066144a14f41f5fdd58ea9219f14c3c9977957d2f8cbbc204695c5beb0},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34091. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Sparse autoencoders (SAEs) decompose dense model activations into discrete latents, making individual features easy to interpret--and easy to misinterpret. In EEG foundation models, this creates a tempting inference: if removing alpha-band activity strongly changes a latent's activation, one might conclude that the latent represents alpha activity. Across 27 settings spanning three backbones, three EEG datasets, and three network depths, this interpretation initially appears compelling: alpha removal changes latent firing 7.3 times more than an equal-width sham notch (95% CI [6.2, 8.7], bootstrapped over settings). However, the alpha filter also deletes far more signal than the sham. After normalizing by removed spectral energy, the ratio falls to 0.28 (95% CI [0.22, 0.36]) and exceeds one in none of the 27 settings. Latents selected for their response to alpha removal are, on clean EEG, slightly anti-correlated with relative alpha power (mean r = -0.073), giving no support for a simple alpha-detector reading. Motivated by this failure case, we propose a validation ladder for semantic interpretations of SAE latents: it asks in turn whether a latent responds, whether that response survives controlling for how much signal the intervention removes, whether it is specific rather than broadly fragile, and whether the proposed property is visible on unperturbed data--while separately testing the stronger claim that the latent matters to a task classifier. Perturbation sensitivity alone does not establish what an SAE latent represents.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.LG).
Categorie arXiv: Machine Learning (cs.LG).
Pagina del lavoro: https://arxiv.org/abs/2609.34091 — 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 2026-10-07T22:14:28.108Z dalla pagina https://arxiv.org/catchup/cs.LG/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).