finding
active
finding:country-token-attends-more-to-the-foreign-city-than-to-any-own-city-98-of-the-time-across-185-sentencesCountry token attends more to the foreign city than to any own city 98% of the time across 185 sentences
Systematic test of same-set suppression.
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Supported by geometry and 98% result but not a full account.
Methods (1)
method
- 185-sentence country-city testsupportsSystematic test of same-set suppression with five cities, one foreign.
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.
- Demonstrates prevalence of token-in-context features and feature splitting of common tokens
- Country query vs own-city key cosine similarity mean −0.47, versus +0.01 for other citiesfinding0.716QK geometry anti-alignment.
- Interesting special case of copying behavior related to tokenization artifacts; primitive precursor to induction heads
- Universalist claim predicting cross-cultural generality.
- Explains hard-to-read dataset patterns.
- Poisoning results.
- §5.5.
- Comparative prediction motivating future work contrasting different approaches to LLM self-knowledge