ASSAY-001: calibration and type-safety evidence on two intent corpora

A frozen-protocol study checks Jev's option probabilities and output schema on Banking77 and CLINC150. It publishes corpus pins, raw responses, scoring rules, amendments, controls and a separate blind rescore.

Notes

Across 8,576 recorded responses, the authors report chosen-option ECE 0.0204 for CLINC150 and 0.0936 for Banking77, using ten equal-width probability bins.

source-backed — Public repository, docs or live artifact. About this label

Destinations

Availability & demo checks

Report a problem →

Filed under

• #benchmark
• #evaluation
• #calibration
• #classification

In collections

• Benchmarks & independent tests →