finding
active
finding:standard-crosscoder-s-identified-features-had-relative-norm-ranks-1-898-qwen-and-6-063-llama-outside-the-top-1-310Standard crosscoder's identified features had relative norm ranks 1,898 (Qwen) and 6,063 (Llama), outside the top 1,310.
Appendix C.4.
Source paper
extracted_from(2026) · Thomas Jiralerspong · Trenton Bricken
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Main architectural claim.
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.
- Footnote 2.
- Figure 5.
- C.5.1.
- Table 1.
- Decoder norm profile across 18 layers.
- Motivation for DFC.
- Cross-judge validation of the primary ESR finding across OpenAI, Alibaba, Anthropic, and Google judge models
- At Llama layer 26 (~81% depth) the predictive component does not dominate the latent norm.finding0.795Fig. 14.