Fetching the paper…

metabench -- A Sparse Benchmark of Reasoning and Knowledge in Large Language Models · Around