Atlas

Models

← All models

Random baseline (EQ-Bench v2)

EQ-Bench · 2024-01-16

`random-baseline` is the non-model chance comparator published on the legacy EQ-Bench v2 leaderboard, where it scores 0.00 on EQ-Bench v2, 25.00 on MAGI-Hard, and 12.50 combined. It is a benchmark-only artifact, not a language model.

Benchmark scores

BenchmarkScore
EQ-Bench v20.0
EQ-Bench v2 + MAGI-Hard Combined12.5
MAGI-Hard25.0
Loading Atlas data…