finding
active
finding:per-token-graph-lacks-any-adversarial-prompt-detection-feature-attribution-flows-to-chemical-compound-names-and-jsonPer-token graph lacks any adversarial prompt detection feature; attribution flows to 'chemical compound names' and 'JSON'
Contrast in §6.3.
Source paper
extracted_from(2026) · Kevin Der · Harish Kamath · Thompson, Ben
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Central 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.
- Structure of the computational pathway.
- Design choice validated empirically.
- §5.5.
- Quantifies method limitations.
- Shows how VPD-identified subnetworks can be analyzed to reveal interpretable pathways of computation (e.g., gender signal routing, syntactic role detection).
- Gradient-based technique using SAE features to estimate causal effects on completions; used to corroborate NLA findings.
- Example A.
- Motivates 14-turn case.