eval.yaml

name: Typed Decisions
description: >
  A benchmark for typed probabilistic decisions. A model gets one piece of
  unstructured state and answers five typed questions about it at once, and
  every answer is a probability distribution, not a single label.

tasks:
  - id: accuracy
    config: all
    split: test
  - id: kl_from_gold
    config: all
    split: test
    lower_is_better: true
  - id: brier
    config: all
    split: test
    lower_is_better: true
  - id: ece
    config: all
    split: test
    lower_is_better: true