Datasets:
Tasks:
Text Classification
Formats:
parquet
Languages:
English
Size:
1K - 10K
Tags:
structured-decisions
calibration
probabilistic-classification
system-one
workflow-evaluation
Synthetic
License:
Submission: SAP RPT 1.6 Decisions API (in-context, 300 examples/workflow): accuracy 0.672
#49
by tomgutt - opened
tomgutt changed discussion title from SAP RPT 1.6 + RPT 1.6 Large: reproducible API evaluation with 300 train examples/workflow to SAP RPT 1.6: reproducible API evaluation with 300 train examples/workflow
Added to the fitted table at #27: 0.672 accuracy, KL 0.646, Brier 0.151, ECE 0.019, 656 ms p50. The Kind column lists it as in-context (300 provider-held train examples per workflow, no weight updates, closed API), and the "Who scored each row" section notes it is your unofficial evaluation of the public Decisions API.
tomgutt changed discussion title from SAP RPT 1.6: reproducible API evaluation with 300 train examples/workflow to Submission: SAP RPT 1.6 Decisions API (in-context, 300 examples/workflow): accuracy 0.672
tomgutt changed discussion status to closed
tomgutt changed discussion status to open
tomgutt changed discussion status to closed