A 42-of-42 failure rate in our LLM benchmark was a parser bug. How we found it, fixed our scorer without fudging results, and ...
Some results have been hidden because they may be inaccessible to you
Show inaccessible resultsSome results have been hidden because they may be inaccessible to you
Show inaccessible results