hypothesis
active
hypothesis:there-are-fewer-representations-competent-for-n-tasks-than-m-n-tasks-so-training-more-general-models-should-yield-fewer-possible-solutionsThere are fewer representations competent for N tasks than M<N tasks, so training more general models should yield fewer possible solutions
Selective pressure toward convergence via task generality
Source paper
extracted_from(2024) · Minyoung Huh · Brian Cheung · Tongzhou Wang · Phillip Isola
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.
- Hypothesis box in §3.2.
- Future direction.
- Bigger models are more likely to converge to a shared representation than smaller modelshypothesis0.819Selective pressure toward convergence via model capacity
- Inverse scaling interpretation.
- Core framing.
- Earlier/less capable models exhibit a larger gap between think and don't think representation strengthfinding0.815Claude 3 models show a bigger difference than newer models like Opus 4.1.
- Fig. 23.
- Claim about current practical feasibility and efficiency of 2-way associative implementations.