Mathematical Reasoning on GSM8K (T, U, F)
91.4T (Success Rate)Binary
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BinaryBackbone=Qwen2.5-7B-Instruct2026.04 | 91.4 | 0 | 8.6 | |
| KARLBackbone=Qwen2.5-7B-Instruct2026.04 | 90.4 | 0.4 | 9.2 | |
| PromptingBackbone=Qwen2.5-7B-Instruct2026.04 | 89.1 | 0.5 | 10.4 | |
| TernaryBackbone=Qwen2.5-7B-Instruct2026.04 | 88.5 | 3 | 8.5 | |
| KARLBackbone=Llama-3.1-8B-Instruct2026.04 | 86.1 | 1.3 | 12.6 | |
| TernaryBackbone=Llama-3.1-8B-Instruct2026.04 | 85.5 | 3 | 11.5 | |
| BinaryBackbone=Llama-3.1-8B-Instruct2026.04 | 85.2 | 0 | 14.8 | |
| PromptingBackbone=Llama-3.1-8B-Instruct2026.04 | 83.7 | 0.5 | 15.8 |