claim
active
claim:turn-averaged-features-generalize-across-span-lengths-even-150x-beyond-training-lengthTurn-averaged features generalize across span lengths, even ~150x beyond training length
Long-context results.
Source paper
extracted_from(2026) · Kevin Der · Harish Kamath · Thompson, Ben
Neighborhood — ranked by edge-count
Papers (1)
paper
Findings (2)
finding
- LongBench v2 example D.
- LongBench v2 example E.
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.
- Prediction from coverage results.
- Figure 1.
- Pairwise coverage.
- 5-way ranking.
- Turn-averaged features have highest embedding similarity 0.663 at k=3; per-token max lowest at 0.617finding0.791Consistent across k=1,2,3,5.
- 20,000 trials.
- Example B.
- Abstract.