v0.19.2: Verifier input format fix
Patch release aligning TrainedVerifier with the production input format documented in the HuggingFace model card. Scores were previously miscalibrated against the tau=0.80 threshold.
Patch release aligning TrainedVerifier with the production input format documented in the HuggingFace model card. Scores were previously miscalibrated against the tau=0.80 threshold.