Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects

Sicherheitsmeldung mit Schweregrad noch nicht bewertet. Technische Details im Tab „Originaltext“; empfohlene Schritte in der Checkliste.

Originaltext der Quelle (nicht übersetzt) – die deutsche Zusammenfassung findest du im Tab „Kurzfassung".

arXiv:2607.20596v2 Announce Type: replace Abstract: Sparse autoencoder (SAE) features are used to interpret and steer large language models, yet nobody has tested whether a feature's causal role is stable across SAE families. Single-token features fire on one vocabulary item, so ground truth permits direct comparison. We analyze 3.9M features across six models and three SAE families and zero-ablate at full layer depth: they sit 4.7x tighter in decoder space and concentrate in early layers. Deleting one lowers the model's logit for that token in 178 of 208 layer conditions, significant after multiple-comparison correction. And depth decides how the damage lands: early-layer deletions disrupt the layers that follow, late-layer deletions change the output directly. Cross-family causal differences exceed within-family scale effects: on the same base model, GemmaScope and BatchTopK features are causally anchored, LlamaScope features locally redundant. Under LlamaScope the token returns to within 2x its pre-ablation rank 96-98% of the time. Changing only the activation function reverses the sign of that difference, so the training recipe is the remaining candidate: cross-family claims are sensitive to training methodology, not just activation function or scale.

Vollständiger Originalartikel ↗

• Betroffenheit im eigenen Stack prüfen: Versionen/Komponenten abgleichen.
• Originalquelle / Hersteller-Advisory lesen: https://arxiv.org/abs/2607.20596
• Verfügbaren Patch oder Workaround einspielen und dokumentieren.