Mathematical Reasoning on GSM8K (test) (Gains, RSE %)
156.22GainCEM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CEMModel=Gemini 2.5 Flash Lite, Template ID=6, K=16, λ=0.12026.05 | 156.22 | 0.32 | |
| CEMModel=Gemini 2.5 Flash Lite, Template ID=6, K=16, λ=0.32026.05 | 85.21 | 0.36 | |
| CEMModel=Gemini 2.5 Flash Lite, Template ID=6, K=16, λ=0.52026.05 | 82.59 | 0.37 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=6, K=16, λ=0.12026.05 | 17.12 | 0.43 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=0, K=16, λ=0.12026.05 | 16.41 | 1.02 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=6, K=8, λ=0.12026.05 | 15.84 | 0.44 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=0, K=16, λ=0.32026.05 | 15.84 | 1.05 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=0, K=16, λ=0.52026.05 | 13.65 | 1.13 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=7, K=16, λ=0.12026.05 | 11.45 | 0.79 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=0, K=8, λ=0.32026.05 | 8.03 | 1.21 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=7, K=8, λ=0.12026.05 | 8.02 | 0.95 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=5, K=16, λ=0.12026.05 | 6.93 | 5.05 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=0, K=8, λ=0.52026.05 | 6.59 | 1.26 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=0, K=8, λ=0.12026.05 | 6.5 | 1.22 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=6, K=4, λ=0.12026.05 | 5.36 | 0.44 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=1, K=16, λ=0.12026.05 | 5.33 | 2.95 | |
| CEMModel=Gemini 2.5 Flash Lite, Template ID=0, K=16, λ=0.12026.05 | 5.29 | 2.35 | |
| CEMModel=gpt-oss-20b-low, Template ID=8, K=24, λ=0.12026.05 | 4.57 | 1.36 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=0, K=4, λ=0.32026.05 | 4.51 | 1.27 | |
| CEMModel=gpt-oss-20b-low, Template ID=8, K=16, λ=0.12026.05 | 4.41 | 1.18 | |
| CEMModel=gpt-oss-20b-low, Template ID=8, K=8, λ=0.12026.05 | 4.36 | 0.84 | |
| CEMModel=Gemini 2.5 Flash Lite, Template ID=0, K=8, λ=0.12026.05 | 4.05 | 1.9 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=0, K=4, λ=0.52026.05 | 3.82 | 1.27 | |
| CEMModel=gpt-oss-20b-low, Template ID=0, K=16, λ=0.12026.05 | 3.81 | 2.95 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=0, K=4, λ=0.12026.05 | 3.68 | 1.24 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=5, K=8, λ=0.12026.05 | 3.67 | 1.77 | |
| CEMModel=gpt-oss-20b-low, Template ID=0, K=24, λ=0.12026.05 | 3.63 | 3.92 | |
| CEMModel=gpt-oss-20b-low, Template ID=0, K=8, λ=0.12026.05 | 3.34 | 2.56 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=7, K=4, λ=0.12026.05 | 3.22 | 1.08 | |
| CEMModel=Gemini 2.5 Flash Lite, Template ID=6, K=8, λ=0.12026.05 | 2.75 | 1.8 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=8, K=4, λ=0.32026.05 | 2.73 | 4.09 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=8, K=4, λ=0.52026.05 | 2.45 | 4.57 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=1, K=8, λ=0.12026.05 | 2.29 | 2.99 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=5, K=4, λ=0.12026.05 | 2.15 | 1.2 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=8, K=4, λ=0.12026.05 | 1.91 | 4.97 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=2, K=16, λ=0.12026.05 | 1.87 | 0.27 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=3, K=16, λ=0.12026.05 | 1.75 | 0.8 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=2, K=4, λ=0.12026.05 | 1.63 | 0.22 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=2, K=8, λ=0.12026.05 | 1.48 | 0.24 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=3, K=8, λ=0.12026.05 | 1.46 | 0.63 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=4, K=8, λ=0.12026.05 | 1.38 | 0.36 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=4, K=4, λ=0.12026.05 | 1.34 | 0.31 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=1, K=4, λ=0.12026.05 | 1.31 | 1.97 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=4, K=16, λ=0.12026.05 | 1.22 | 0.4 | |
| CEMModel=Qwen2.5-Math-7B-Instruct, Template ID=3, K=4, λ=0.12026.05 | 1.07 | 0.43 | |
| CEMModel=Gemini 2.5 Flash Lite, Template ID=6, K=4, λ=0.12026.05 | 1.06 | 0.99 | |
| CEMModel=Gemini 2.5 Flash Lite, Template ID=0, K=4, λ=0.12026.05 | 0.96 | 1.95 |