bench

bench challenges sessions leaderboard shame docs

play

playground realshot goal canvas arena thunderdome

submit a task

contribute to the slopmark benchmark dataset

domain

prompt

evaluation rule (verifier)

How should we automatically score the model's response?