Successive-answer demonstrations induce repeated switches between continuing the answer sequence and solving the actual test question in OLMo3 and Apertus. The larger OLMo3 model generalizes more often but can collapse abruptly.

Successive-answer demonstrations induce repeated switches between continuing the answer sequence and solving the actual test question in OLMo3 and Apertus. The larger OLMo3 model generalizes more often but can collapse abruptly. · CiteArk

Measurement results

0/3 assessed

0 supported0 disputed or conflicting3 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
model: OLMo3-32B · dataset: Arithmetic (+1) · pretrainingTokensTrillions: 2.17	81%	—	Not assessed
model: OLMo3-32B · dataset: Arithmetic (+1) · pretrainingTokensTrillions: 2.19	0%	—	Not assessed
model: OLMo3-32B · dataset: Arithmetic (+1) · pretrainingTokensTrillions: 2.21	81.7%	—	Not assessed

Successive-answer demonstrations induce repeated switches between continuing the answer sequence and solving the actual test question in OLMo3 and Apertus. The larger OLMo3 model generalizes more often but can collapse abruptly.

Reported and observed measurements

Assessments (0)

Runs (0)

Measurement results