Math Reasoning on AIME 25 (Acc. (%), ∆ (%))
20Accuracy (%)ETO
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| ETOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=true2026.05 | 20 | 13.3 | |
| ED-iDPOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=true2026.05 | 20 | 13.3 | |
| ED-GRPOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=true2026.05 | 20 | 13.3 | |
| GRPOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=true2026.05 | 16.7 | 10 | |
| DAPOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=true2026.05 | 16.7 | 10 | |
| ED-GRPOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=false2026.05 | 16.7 | 10 | |
| CoTBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=true2026.05 | 13.3 | 6.6 | |
| GRPOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=false2026.05 | 13.3 | 6.6 | |
| ED-iDPOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=false2026.05 | 13.3 | 6.6 | |
| ETOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=false2026.05 | 10 | 3.3 | |
| DAPOBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=false2026.05 | 10 | 3.3 | |
| CoTBase Model=Qwen2.5-7B-Instruct-SFT, Self-Consistency=false2026.05 | 6.7 | — | |
| ED-GRPOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=false2026.05 | 6.7 | 6.7 | |
| ED-GRPOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=true2026.05 | 6.7 | 6.7 | |
| CoTBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=true2026.05 | 3.3 | 3.3 | |
| ETOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=false2026.05 | 3.3 | 3.3 | |
| ETOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=true2026.05 | 3.3 | 3.3 | |
| ED-iDPOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=false2026.05 | 3.3 | 3.3 | |
| ED-iDPOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=true2026.05 | 3.3 | 3.3 | |
| CoTBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=false2026.05 | 0 | — | |
| GRPOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=false2026.05 | 0 | — | |
| GRPOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=true2026.05 | 0 | — | |
| DAPOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=false2026.05 | 0 | — | |
| DAPOBase Model=LLaMA3.1-8B-Instruct-SFT, Self-Consistency=true2026.05 | 0 | — |