default_rlm_best_stripped — AIME 2025 prompt evaluation results from GEPA prompt optimization. Evaluated on 150 AIME 2025 problems (MathArena/aime_2025 x5) with 10 samples per problem using openai/gpt-4.1-mini at temperature 0.7.

	Results

Metric Value

pass@1 0.3800

pass@2 0.4667

pass@3 0.4933

pass@4 0.5267

pass@5 0.5533

pass@6 0.5867

pass@7 0.5933

pass@8 0.6133

pass@9 0.6133

pass@10 0.6400… See the full description on the dataset page: https://huggingface.co/datasets/reasoning-degeneration-dev/aime-2025-gepa-default-rlm-best-stripped.