| --- |
|
|
| ## license: mit |
|
|
|
|
| | 配置 | GSM8K | SVAMP | GSM8K-hard | MultiArith | |
| | -------------------------- | ---------- | ---------- | ---------- | ---------- | |
| | **zero_shot** | 74.15% | 81.00% | 35.86% | 99.44% | |
| | eval_full_s0.0015_a0.0002 | 73.84% | 81.33% | 36.62% | 99.44% | |
| | eval_full_s0.0015_a0.00025 | 73.16% | 83.00% | 37.45% | 97.22% | |
| | **avg_full** | 73.50% | 82.17% | 37.04% | 98.50% | |
| | eval_r1_s0.045_a0.0025 | 74.60% | 83.00% | 37.30% | 98.33% | |
| | eval_r1_s0.045_a0.003 | 74.91% | 85.67% | 37.45% | 98.33% | |
| | eval_r1_s0.04_a0.002 | 75.28% | 83.00% | 38.29% | 97.78% | |
| | eval_r4_s0.02_a0.005 | 75.97% | 85.33% | 39.35% | 100.00% | |
| | eval_r4_s0.02_a0.0075 | 76.65% | 83.33% | 36.39% | 96.67% | |
| | eval_r4_s0.025_a0.0025 | 75.36% | 85.33% | 38.06% | 97.22% | |
| | **avg_r4** | 75.99% | 84.66% | 37.93% | 97.96% | |
| | eval_r8_s0.0225_a0.0035 | 75.66% | 83.00% | 37.98% | 98.33% | |
| | eval_r8_s0.0225_a0.0025 | 76.95% | 84.33% | 38.67% | 99.44% | |
| | eval_r8_s0.0225_a0.0025 | 76.42% | 82.33% | 37.98% | 97.78% | |
| | **avg_r8** | 76.34% | 83.22% | 38.21% | 98.52% | |
| | eval_r16_s0.0225_a0.005 | 75.82% | 85.67% | 39.12% | 98.89% | |
| | eval_r16_s0.0225_a0.0025 | 74.83% | 85.00% | 37.83% | 96.67% | |
| | eval_r16_s0.0225_a0.0035 | 75.66% | 83.67% | 38.89% | 98.33% | |
| | **avg_r16** | 75.44% | 84.78% | **38.61%** | 97.96% | |
| | eval_r64_s0.01_a0.0025 | 77.03% | 83.33% | 36.62% | 96.11% | |
| | eval_r64_s0.0085_a0.0025 | 75.59% | 84.33% | 37.68% | 96.11% | |
| | eval_r128_s0.008_a0.002 | 76.19% | 85.33% | 38.36% | 99.44% | |
| | eval_r128_s0.0075_a0.002 | 75.74% | 85.00% | 38.29% | 98.89% | |
| | eval_r128_s0.0075_a0.00225 | 77.48% | 85.00% | 37.45% | 100.00% | |
| | **avg_r128** | **76.47%** | **85.11%** | 38.03% | **99.44%** | |
| | eval_r256_s0.00575_a0.00135 | 76.04% | 86.67% | 37.83% | 98.33% | |
| | eval_r256_s0.00575_a0.0015 | 77.03% | 84.67% | 36.24% | 97.22% | |
| | eval_r256_s0.006_a0.00135 | 77.10% | 84.33% | 37.53% | 98.89% | |
| | eval_r512_s0.004_a0.0012 | 75.82% | 81.00% | 35.33% | 97.22% | |
| | eval_r512_s0.00425_a0.001 | 76.35% | 81.67% | 36.01% | 98.89% | |
| | eval_r512_s0.00425_a0.00125 | 74.83% | 86.67% | 35.03% | 98.89% | |
| |
| |
| |