finding
active
finding:at-one-instance-per-batch-2-4e-7-larger-models-reach-lower-task-loss-at-the-same-compute-smaller-models-plateauAt one instance per batch (2.4e-7), larger models reach lower task loss at the same compute; smaller models plateau
Fig. 23.
Source paper
extracted_from(2026) · Jing Huang · Daniel Wurgaft · Rachit Bansal · Laura Ruis +5
Neighborhood — ranked by edge-count
Papers (1)
paper
Claims (1)
claim
- Core phenomenological argument.
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.
- Fig. 22.
- Hypothesis box in §3.2.
- Inverse scaling interpretation.
- Selective pressure toward convergence via task generality
- Bigger models are more likely to converge to a shared representation than smaller modelshypothesis0.804Selective pressure toward convergence via model capacity
- Core framing.
- Future work for frequent tasks.
- Extrapolation of scaling predictive models to AGI.