ginigen-ai's picture
README: correct model count 22 -> 33 (matches leaderboard_mcq.json)
9832311 verified
|
Raw
History Blame Contribute Delete
877 Bytes
metadata
title: Metacognition Leaderboard
emoji: 🧠
colorFrom: indigo
colorTo: pink
sdk: docker
app_port: 7860
pinned: false
license: apache-2.0

🧠 Metacognition Leaderboard

Live, visual leaderboard for ginigen-ai/Metacognition-Bench.

33 LLMs ranked by metacognitive vulnerability (trap_rate): given a 4-choice version of each trap problem (correct option + a seductive trap distractor), how often the model fails to pick the correct answer. Lower = stronger metacognition.