finding
active
finding:clt-heatmaps-show-broad-similarity-to-hanna-et-al-in-columnar-structure-but-less-input-dependence-for-top-featuresCLT heatmaps show broad similarity to Hanna et al. in columnar structure but less input dependence for top features
Comparison of heatmaps.
Source paper
extracted_fromNeighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Difference between sparse and neuron analyses.
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.
- Main conclusion.
- Speculative explanation.
- How do CLT number representations relate to those in non-sparse models, such as helical representations?question0.750Future work.
- Feature 425104 heatmap promotes outputs above ~60 with little input dependence, e.g. promotes 62 given 90finding0.743Diverges from Hanna et al. heatmaps.
- Baseline.
- Third empirical pattern.
- Input-output mapping per feature across ~100 prompts.
- Key advantage statement.