nm-research's picture
Update BFCL/SWE-Bench eval table + add recovery-vs-base-model comparison for reasoning benchmarks
317300b verified