Benchmark
Claw-Eval
A benchmark designed to evaluate the capabilities of code and reasoning agents, commonly used as an out-of-distribution generalization test set.
Timeline (last 90 days)
Oct 1
The second stage continued evolving on the out-of-distribution benchmark Claw-Eval, raising the score from 66.17 to 68.06, surpassing Codex.
Unverified50%