finding
active
finding:sae-based-filtering-and-reward-shaping-show-lower-off-target-effects-but-lower-overall-reductionSAE-based filtering and reward shaping show lower off-target effects but lower overall reduction
Fig. 9.
Source paper
extracted_from(2026) · Leon Bergen · Usha Bhalla · Sidharth Baskaran · Max Loeffler +13
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Hypothesized explanation.
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.
- Safeguards analysis.
- Extension of mechanistic interpretability findings to the metacognitive domain
- Robustness influences superposition, predicting lower SAE loss in more robust models.hypothesis0.775Popperian credit for surprising prediction.
- Policy recommendation based on latent #10 activating at 5% incorrect data before behavioral misalignment emerges
- Explains why variance correction is needed to see the self-evaluation–persistence relationship
- A promising property for interpretability analysis off-distribution.
- Text-based and self-steered emotionality ratings for SAE features are correlated at only ρ = +0.051 (n.s.).finding0.766Shows low agreement between the two evaluation modalities
- If agentic self-steering evaluation proves robust, it might be used to better explain and interpret SAE features in generalhypothesis0.762Speculative claim about scaling introspective access to general SAE feature interpretation