finding
active
finding:saes-trained-on-fineweb-or-a-non-chat-model-produced-interesting-features-but-transferred-less-well-to-chat-applicationsSAEs trained on FineWeb or a non-chat model produced interesting features but transferred less well to chat applications
Comparison against LMSYS training.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Hypotheses (1)
hypothesis
- Explanation for LMSYS superiority.
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.
- Feature cluster examples from the UMAP.
- SAE latents depend on training data.
- Extension of mechanistic interpretability findings to the metacognitive domain
- Out-of-distribution generalization of SAE features.
- App. B.
- A promising property for interpretability analysis off-distribution.
- Including chat data in SAE training merges RM bias features with generic Assistant featureshypothesis0.752Explains poor features with three-dataset SAE.
- Prediction following diversity-undertraining logic.