Formal Instruction Generation on Deck
119.87Total Time (s)MAFIG
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| MAFIGbackbone=Qwen2.5-Coder-7B2026.04 | 119.87 | 0.19 | 97.5 | |
| Qwen2.5-Coder-7B2026.04 | 250.13 | 0.39 | 17.45 | |
| Qwen3-Coder-480B-A35B-Instruct2026.04 | 2,646.4 | 4.12 | 51.4 | |
| Qwen3-32B2026.04 | 8,300.6 | 12.93 | 50.62 | |
| DeepSeek-V3.22026.04 | 8,753.61 | 13.63 | 50.93 | |
| GLM-4.72026.04 | 20,582.86 | 32.06 | 83.49 |