Datasets:
Tasks:
Text Classification
Formats:
parquet
Languages:
English
Size:
1K - 10K
Tags:
structured-decisions
calibration
probabilistic-classification
system-one
workflow-evaluation
Synthetic
License:
Add Amos 1 medium and large (general, zero-shot) and the large's fitted row
#32
by mukaddamzaid - opened
Adds the rows from discussion #31: Amos 1 medium 0.7645 / KL 0.151 and Amos 1 large 0.761 / KL 0.126 to the general table (self-reported, marked §), and the large with per-question priors fitted on train (0.8025 / KL 0.122, no weights changed) to the second table. Both models can be rescored at https://api.amos.scira.ai/v1/systemone (no key), model amos-1-medium or amos-1-large.
Merged, thanks. Amos 1 medium and large are in the zero-shot table and the large's fitted row is in the second table.
codelion changed pull request status to merged