concept
active
concept:relevance-filtering-of-sae-latentsRelevance Filtering of SAE Latents
Pre-filtering step excluding latents naturally activated by each prompt to ensure genuine off-topic steering
Neighborhood — ranked by edge-count
Concepts (2)
concept
- SAE Latentsrelated_toInterpretable features extracted by sparse autoencoders used as steering targets in this study
- Endogenous Steering ResistancesupportsThe central phenomenon introduced by this paper: inference-time recovery from irrelevant activation steering in LLMs
Related by similarity (8)
cosine ≥ 0.65 · no typed edgeEntities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.
- Pre-filtering step excluding abstract latents where off-topic detection is harder
- Using GPT-4o or o3 to automatically generate interpretations of SAE latents from top-activating examples
- Standard baseline SAE operating on individual token activations.
- Reduces SAE optimization to allocating N latents across features.
- Adding a multiple of the SAE latent decoder vector to token activations to causally test each latent's role in misalignment
- Interpretability method criticized in this paper for shattering manifolds into atomic pieces, obscuring overarching semantic structure.
- Interpretability tool underlying top-feature, details, steering tools.
- Backprop of logit-minus-mean-logit loss to surface causal intervention candidates, summed across token positions.