Reasoning on AIME24
81.5AccuracyParallel-Probe
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Parallel-ProbeBase Model=Qwen3-8B, Type=Parallel2026.02 | 81.5 | 20,300 | 730,800 | |
| SC@64Base Model=Qwen3-8B, Type=Parallel2026.02 | 80.4 | 30,100 | 910,800 | |
| ASCBase Model=Qwen3-8B, Type=Seq.2026.02 | 80.4 | 226,000 | 226,000 | |
| ESCBase Model=Qwen3-8B, Type=Hybrid2026.02 | 80.4 | 84,700 | 459,400 | |
| SC@64Base Model=Qwen3-4B, Type=Parallel2026.02 | 80 | 29,300 | 886,800 | |
| ASCBase Model=Qwen3-4B, Type=Seq.2026.02 | 80 | 214,200 | 214,200 | |
| ESCBase Model=Qwen3-4B, Type=Hybrid2026.02 | 80 | 98,900 | 528,900 | |
| SC@64 + SACBase Model=Qwen3-4B, Type=Parallel2026.02 | 80 | 24,800 | 782,200 | |
| Parallel-ProbeBase Model=Qwen3-4B, Type=Parallel2026.02 | 79.7 | 19,200 | 688,900 | |
| SC@64 + SACBase Model=Qwen3-8B, Type=Parallel2026.02 | 76.7 | 25,600 | 773,400 | |
| SC@64Base Model=Qwen3-1.7B, Type=Parallel2026.02 | 72.5 | 31,400 | 1,025,800 | |
| ESCBase Model=Qwen3-1.7B, Type=Hybrid2026.02 | 72.5 | 170,400 | 909,200 | |
| ASCBase Model=Qwen3-1.7B, Type=Seq.2026.02 | 72.3 | 482,600 | 482,600 | |
| Parallel-ProbeBase Model=Qwen3-1.7B, Type=Parallel2026.02 | 68.1 | 20,500 | 748,500 | |
| SC@64 + SACBase Model=Qwen3-1.7B, Type=Parallel2026.02 | 64.5 | 27,300 | 868,200 | |
| Qwen3-4B + TTSVModel=Qwen3-4B, Adaptation=TTSV2025.12 | 60 | — | — | |
| Qwen3-4BModel=Qwen3-4B2025.12 | 56.67 | — | — | |
| Parallel-ProbeBase Model=Qwen3-0.6B, Type=Parallel2026.02 | 21.8 | 20,800 | 773,800 | |
| SC@64Base Model=Qwen3-0.6B, Type=Parallel2026.02 | 21.4 | 32,700 | 1,008,600 | |
| ASCBase Model=Qwen3-0.6B, Type=Seq.2026.02 | 21.4 | 805,500 | 805,500 | |
| ESCBase Model=Qwen3-0.6B, Type=Hybrid2026.02 | 21.4 | 192,900 | 986,700 | |
| SC@64 + SACBase Model=Qwen3-0.6B, Type=Parallel2026.02 | 19.5 | 26,800 | 820,700 |