Mathematical Reasoning on GSM8K (ECE, BS)
4.28ECECaliDist (As.)
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CaliDist (As.)Backbone=QWEN-3 8B2026.06 | 4.28 | 7.16 | |
| CaliDist (As.)Backbone=LLAMA-3.1 8B2026.06 | 5.51 | 14.13 | |
| ConsistencyBackbone=LLAMA-3.1 8B2026.06 | 6.12 | 7.83 | |
| FSDBackbone=QWEN-3 8B2026.06 | 6.18 | 6.53 | |
| EntropyBackbone=QWEN-3 8B2026.06 | 6.78 | 6.75 | |
| ConsistencyBackbone=QWEN-3 8B2026.06 | 6.97 | 6.91 | |
| VanillaBackbone=QWEN-3 8B2026.06 | 10.48 | 10.48 | |
| FSDBackbone=LLAMA-3.1 8B2026.06 | 13.19 | 11.73 | |
| VanillaBackbone=LLAMA-3.1 8B2026.06 | 16.71 | 16.8 | |
| EntropyBackbone=LLAMA-3.1 8B2026.06 | 21.13 | 16.7 |