mirror of
https://github.com/hanzoai/benchmarks.git
synced 2026-10-30 14:59:11 +00:00
Copied from hanzoai/decision results/laya/parity.json: max logit difference 2.9e-5 and max probability difference 6.2e-6 against a 1e-4 bound, across nine behaviours and a twelve-language set, on all three checkpoints. Co-authored-by: Hanzo Dev <[email protected]>