Comprehensive AIME 2025 prompt evaluation results from GEPA (Genetic-Pareto) prompt optimization experiments. Each prompt was evaluated on the full 150-example AIME 2025 test set (MathArena/aime_2025 x5) with 10 samples per problem using openai/gpt-4.1-mini at temperature 0.7.

	Results (sorted by pass@10)

Prompt pass@1 pass@3 pass@5 pass@10

best_vanilla 0.3667 0.4933 0.5533 0.6267

inputs_only_vanilla_final 0.4267 0.5000 0.5800… See the full description on the dataset page: https://huggingface.co/datasets/reasoning-degeneration-dev/aime-2025-gepa-prompt-eval-k10.