After matched math SFT, the selected 4.5T OLMo3-32B checkpoint generalizes better to GPQA than the 4.9T checkpoint (36.3% versus 29.8%). The wider checkpoint comparison shows its strongest GPQA gain despite further pretraining/mid-training improving in-distribution performance.

After matched math SFT, the selected 4.5T OLMo3-32B checkpoint generalizes better to GPQA than the 4.9T checkpoint (36.3% versus 29.8%). The wider checkpoint comparison shows its strongest GPQA gain despite further pretraining/mid-training improving in-distribution performance. · CiteArk

Measurement results

0/2 assessed

0 supported0 disputed or conflicting2 not assessed

The complete result group stays with its claim. Assessed rows come first; their judgments do not replace the whole-claim conclusion. Expand a row for conditions, sources, and evidence history.

Each row is a measurement with its reported value, latest available observation, and cumulative evidence state.
Metric and conditions	Reported	Latest observation	Evidence judgment
dataset: GPQA-Diamond · posttraining: Math SFT · basePretrainingTokensTrillions: 4.5	36.3%	—	Not assessed
dataset: GPQA-Diamond · posttraining: Math SFT · basePretrainingTokensTrillions: 4.9	29.8%	—	Not assessed

After matched math SFT, the selected 4.5T OLMo3-32B checkpoint generalizes better to GPQA than the 4.9T checkpoint (36.3% versus 29.8%). The wider checkpoint comparison shows its strongest GPQA gain despite further pretraining/mid-training improving in-distribution performance.

Reported and observed measurements

Assessments (0)

Runs (0)

Measurement results