Mathematical Reasoning on AIME24, AIME25, HMMT25, Minerva
76.4AIME 2024 ScoreTP-Schedule (8B)
Evaluation Results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| TP-Schedule (8B)Backbone=Qwen3-8B, Credit Allocation Method=GRPO, Temporal Scheduling=true, Trajectory Percentile (TP)=true2026.05 | 76.4 | 70 | 44.3 | 84.5 | 68.8 | |
| GRPO (8B)Backbone=Qwen3-8B, Credit Allocation Method=GRPO2026.05 | 75.6 | 68.6 | 42.7 | 83.6 | 68 | |
| Qwen3-8BBackbone=Qwen3-8B2026.05 | 74.8 | 66.3 | 41.8 | 82.3 | 66.3 | |
| Entropy Adv. + ScheduleBackbone=Qwen3-4B, Credit Allocation Method=Entropy Adv., Temporal Scheduling=true2026.05 | 74.4 | 66.7 | 42.4 | 80.1 | 65.4 | |
| TP-ScheduleBackbone=Qwen3-4B, Credit Allocation Method=GRPO, Temporal Scheduling=true, Trajectory Percentile (TP)=true2026.05 | 74.2 | 67.1 | 44.8 | 83.8 | 67.5 | |
| Forking Tok. + ScheduleBackbone=Qwen3-4B, Credit Allocation Method=Forking Tok., Temporal Scheduling=true2026.05 | 74.2 | 65.8 | 42.7 | 82.1 | 66.2 | |
| Entropy Adv.Backbone=Qwen3-4B, Credit Allocation Method=Entropy Adv.2026.05 | 73.8 | 65.8 | 41.9 | 80.1 | 65.4 | |
| GRPOBackbone=Qwen3-4B, Credit Allocation Method=GRPO2026.05 | 73.1 | 65.3 | 41.7 | 80.9 | 65.3 | |
| TP-Schedule (GSPO)Backbone=Qwen3-4B, Credit Allocation Method=GSPO, Temporal Scheduling=true, Trajectory Percentile (TP)=true2026.05 | 72.3 | 66.8 | 44 | 80.2 | 65.8 | |
| GSPOBackbone=Qwen3-4B, Credit Allocation Method=GSPO2026.05 | 71.5 | 65.6 | 43.3 | 79.8 | 65 | |
| Forking Tok.Backbone=Qwen3-4B, Credit Allocation Method=Forking Tok.2026.05 | 71.5 | 64.8 | 41.5 | 81.3 | 64.8 | |
| Qwen3-4BBackbone=Qwen3-4B2026.05 | 71.2 | 62.9 | 40 | 79.8 | 63.5 |