Mathematical Reasoning on AMC 2025 (test)
76.31Acc@5LARK
Evaluation Results
| Method | Links | |
|---|---|---|
| LARKBackbone=Qwen-2.5-7B, Selection Budget (B)=32026.05 | 76.31 | |
| Rule-based QualitymaxBackbone=Qwen-2.5-7B, Selection Budget (B)=32026.05 | 67.07 | |
| RSRminBackbone=Qwen-2.5-7B, Selection Budget (B)=32026.05 | 66.27 | |
| LARKBackbone=Qwen-2.5-7B, Selection Budget (B)=12026.05 | 65.86 | |
| Local NaturalnessmaxBackbone=Qwen-2.5-7B, Selection Budget (B)=32026.05 | 64.26 | |
| GRAPEmaxBackbone=Qwen-2.5-7B, Selection Budget (B)=32026.05 | 63.05 | |
| LLM-judged QualitymaxBackbone=Qwen-2.5-7B, Selection Budget (B)=32026.05 | 62.25 | |
| RandomBackbone=Qwen-2.5-7B, Selection Budget (B)=32026.05 | 61.45 | |
| RSRminBackbone=Qwen-2.5-7B, Selection Budget (B)=12026.05 | 59.44 | |
| Local NaturalnessmaxBackbone=Qwen-2.5-7B, Selection Budget (B)=12026.05 | 57.83 | |
| Token LengthmaxBackbone=Qwen-2.5-7B, Selection Budget (B)=32026.05 | 57.03 | |
| Rule-based QualitymaxBackbone=Qwen-2.5-7B, Selection Budget (B)=12026.05 | 56.63 | |
| LLM-judged QualitymaxBackbone=Qwen-2.5-7B, Selection Budget (B)=12026.05 | 56.22 | |
| GRAPEmaxBackbone=Qwen-2.5-7B, Selection Budget (B)=12026.05 | 55.02 | |
| RandomBackbone=Qwen-2.5-7B, Selection Budget (B)=12026.05 | 51 | |
| Token LengthmaxBackbone=Qwen-2.5-7B, Selection Budget (B)=12026.05 | 50.6 | |
| BaseModelBackbone=Qwen-2.5-7B, Selection Budget (B)=12026.05 | 47.79 | |
| BaseModelBackbone=Qwen-2.5-7B, Selection Budget (B)=32026.05 | 47.79 | |
| LARKBackbone=Qwen-2.5-1.5B, Selection Budget (B)=32026.05 | 42.57 | |
| LARKBackbone=Qwen-2.5-1.5B, Selection Budget (B)=12026.05 | 40.56 | |
| Local NaturalnessmaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=32026.05 | 38.96 | |
| Rule-based QualitymaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=32026.05 | 38.55 | |
| GRAPEmaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=32026.05 | 38.15 | |
| Local NaturalnessmaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=12026.05 | 36.14 | |
| RSRminBackbone=Qwen-2.5-1.5B, Selection Budget (B)=32026.05 | 36.14 | |
| GRAPEmaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=12026.05 | 35.34 | |
| LLM-judged QualitymaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=32026.05 | 35.34 | |
| RandomBackbone=Qwen-2.5-1.5B, Selection Budget (B)=12026.05 | 33.33 | |
| RandomBackbone=Qwen-2.5-1.5B, Selection Budget (B)=32026.05 | 32.93 | |
| RSRminBackbone=Qwen-2.5-1.5B, Selection Budget (B)=12026.05 | 32.13 | |
| Rule-based QualitymaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=12026.05 | 29.72 | |
| LLM-judged QualitymaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=12026.05 | 29.32 | |
| Token LengthmaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=32026.05 | 25.7 | |
| Token LengthmaxBackbone=Qwen-2.5-1.5B, Selection Budget (B)=12026.05 | 25.3 | |
| BaseModelBackbone=Qwen-2.5-1.5B, Selection Budget (B)=12026.05 | 24.5 | |
| BaseModelBackbone=Qwen-2.5-1.5B, Selection Budget (B)=32026.05 | 24.5 | |
| LARKBackbone=Llama-3.2-3B, Selection Budget (B)=32026.05 | 22.09 | |
| GRAPEmaxBackbone=Llama-3.2-3B, Selection Budget (B)=32026.05 | 20.48 | |
| Local NaturalnessmaxBackbone=Llama-3.2-3B, Selection Budget (B)=32026.05 | 20.08 | |
| LARKBackbone=Llama-3.2-3B, Selection Budget (B)=12026.05 | 19.28 | |
| LLM-judged QualitymaxBackbone=Llama-3.2-3B, Selection Budget (B)=12026.05 | 18.88 | |
| LLM-judged QualitymaxBackbone=Llama-3.2-3B, Selection Budget (B)=32026.05 | 18.47 | |
| Local NaturalnessmaxBackbone=Llama-3.2-3B, Selection Budget (B)=12026.05 | 17.67 | |
| RandomBackbone=Llama-3.2-3B, Selection Budget (B)=32026.05 | 17.67 | |
| RSRminBackbone=Llama-3.2-3B, Selection Budget (B)=32026.05 | 17.27 | |
| Token LengthmaxBackbone=Llama-3.2-3B, Selection Budget (B)=12026.05 | 16.47 | |
| Token LengthmaxBackbone=Llama-3.2-3B, Selection Budget (B)=32026.05 | 16.06 | |
| RSRminBackbone=Llama-3.2-3B, Selection Budget (B)=12026.05 | 15.66 | |
| Rule-based QualitymaxBackbone=Llama-3.2-3B, Selection Budget (B)=32026.05 | 15.66 | |
| RandomBackbone=Llama-3.2-3B, Selection Budget (B)=12026.05 | 14.86 | |
| GRAPEmaxBackbone=Llama-3.2-3B, Selection Budget (B)=12026.05 | 14.46 | |
| BaseModelBackbone=Llama-3.2-3B, Selection Budget (B)=12026.05 | 11.24 | |
| BaseModelBackbone=Llama-3.2-3B, Selection Budget (B)=32026.05 | 11.24 | |
| Rule-based QualitymaxBackbone=Llama-3.2-3B, Selection Budget (B)=12026.05 | 10.44 |