.eval_results/swe_bench_verified.yaml

- dataset:
    id: SWE-bench/SWE-bench_Verified
    task_id: swe_bench_%_resolved
  value: 70.3
  source:
    url: https://huggingface.co/OrionLLM/GRM-3.2-Cliff
    name: Official GRM-3.2 Benchmark
    user: DedeProGames