eval.yaml

name: FINAL Bench
description: >
  Measuring Functional Metacognitive Reasoning in Large Language Models.
  100 expert-level tasks with hidden cognitive traps across 15 domains
  and 8 TICOS metacognitive types.
evaluation_framework: final-bench

tasks:
  - id: metacognitive
    config: default
    split: train