.eval_results/mellum2.yaml

# Mellum 2 Thinking — evaluation results
# Self-reported by JetBrains. Post-training evaluations from Mellum 2 Technical Report
# (Table 10, RL column).
# Only entries for benchmarks confirmed to be registered as HF Hub Benchmarks are listed
# (a Benchmark is a dataset repo with `eval.yaml` at its root).

- dataset:
    id: Idavidrein/gpqa
    task_id: diamond
  value: 57.6
  date: "2026-05-27"
  notes: "post-training eval (SFT + RLVR), with thinking, no-tools"

- dataset:
    id: gorilla-llm/Berkeley-Function-Calling-Leaderboard
    task_id: bfclv3
  value: 69.4
  date: "2026-05-27"
  notes: "post-training eval (SFT + RLVR), with thinking, tools"