finding
active
finding:a-2x-active-feature-multiplier-gave-best-concept-recovery-for-dsf-crosscoderA 2x active-feature multiplier gave best concept recovery for DSF crosscoder.
Hyperparameter sweep.
Source paper
extracted_from(2026) · Thomas Jiralerspong · Trenton Bricken
Neighborhood — ranked by edge-count
Papers (1)
paper
Frameworks (1)
framework
- Baseline from Mishra-Sharma et al. dedicating features to shared variance.
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.
- Figure 4 left.
- Figure 5.
- Motivation for DFC.
- Table 1.
- Decoder norm profile across 18 layers.
- Crosscoder model diffing is best used as a high-recall pre-screening tool followed by verification.claim0.763Recommended workflow.
- Architectural variant partitioning the dictionary into A-exclusive, B-exclusive and shared sets.
- Compared with base SAE transfer.