steer-full_3b / run /capability /capability_summary.csv

run directory (provenance + evals)

7f4d889 verified about 1 month ago

128 Bytes

	benchmark,metric,m0,m1,delta_pp
	mmlu,"acc,none",67.72,68.42,+0.70
	gsm8k_cot,"exact_match,flexible-extract",61.00,68.00,+7.00