List your company

Mathematical Reasoning on AIME 2025 (Pass@1 Accuracy, Training Time, Training Time Reduction)

33.5Pass@1 AccuracyPRUNE-OPD (overlap)
1.6769.93818.226.462May 8, 2026

Evaluation results

MethodLinks
PRUNE-OPD (overlap)Student-Teacher Pair=DeepSeek-R1-Distill-Qwen-1.5B / JustRL-DeepSeek-1.5B2026.0533.54.140.6
OPDStudent-Teacher Pair=DeepSeek-R1-Distill-Qwen-1.5B / JustRL-DeepSeek-1.5B2026.0532.76.90
OPD (Truncate 4k)Student-Teacher Pair=DeepSeek-R1-Distill-Qwen-1.5B / JustRL-DeepSeek-1.5B2026.05314.830.4
PRUNE-OPD (top-p)Student-Teacher Pair=DeepSeek-R1-Distill-Qwen-1.5B / JustRL-DeepSeek-1.5B2026.0530.84.337.7
OPD (Truncate 4k)Student-Teacher Pair=DeepSeek-R1-Distill-Qwen-1.5B / DeepSeek-R1-Distill-Qwen-7B2026.0525.86.647.2
PRUNE-OPD (overlap)Student-Teacher Pair=DeepSeek-R1-Distill-Qwen-1.5B / DeepSeek-R1-Distill-Qwen-7B2026.0525.8468
OPDStudent-Teacher Pair=DeepSeek-R1-Distill-Qwen-1.5B / DeepSeek-R1-Distill-Qwen-7B2026.0525.412.50
PRUNE-OPD (top-p)Student-Teacher Pair=DeepSeek-R1-Distill-Qwen-1.5B / DeepSeek-R1-Distill-Qwen-7B2026.0524.24.861.6
PRUNE-OPD (overlap)Student-Teacher Pair=Qwen3-4B-Base / Qwen3-4B (Non-thinking)2026.0514.613.552.6
OPD (Truncate 4k)Student-Teacher Pair=Qwen3-4B-Base / Qwen3-4B (Non-thinking)2026.0513.522.321.8
PRUNE-OPD (top-p)Student-Teacher Pair=Qwen3-4B-Base / Qwen3-4B (Non-thinking)2026.0512.72029.8
OPDStudent-Teacher Pair=Qwen3-4B-Base / Qwen3-4B (Non-thinking)2026.0510.628.50
PRUNE-OPD (overlap)Student-Teacher Pair=Qwen3-1.7B-Base / Qwen3-4B (Non-thinking)2026.054.210.337.6
OPDStudent-Teacher Pair=Qwen3-1.7B-Base / Qwen3-4B (Non-thinking)2026.052.916.50
OPD (Truncate 4k)Student-Teacher Pair=Qwen3-1.7B-Base / Qwen3-4B (Non-thinking)2026.052.912.921.8
PRUNE-OPD (top-p)Student-Teacher Pair=Qwen3-1.7B-Base / Qwen3-4B (Non-thinking)2026.052.911.530.3