finding
active
finding:at-equal-general-lm-loss-larger-models-achieve-lower-task-loss-when-frequency-is-low-2-4e-7-similar-trajectories-at-higher-frequencyAt equal general LM loss, larger models achieve lower task loss when frequency is low (2.4e-7); similar trajectories at higher frequency
Fig. 22.
Source paper
extracted_from(2026) · Jing Huang · Daniel Wurgaft · Rachit Bansal · Laura Ruis +5
Neighborhood — ranked by edge-count
Papers (1)
paper
Hypotheses (1)
hypothesis
- Hypothesis box in §3.2.
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. 23.
- Larger OLMo models achieve both low training loss and high test accuracy, i.e. learn generalizable structurefinding0.814Fig. 5.
- Data-induced competition over resources.
- Suspicion from energy level jumps.
- Practical implication.
- Selective pressure toward convergence via task generality
- Fig. 5.
- Baseline.