Formal Language Generation on JSON-Bench
99.7Syntactic Accuracy @1LAVE
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| LAVEModel backbone=LLaDA-1.52026.01 | 99.7 | 99.7 | 99.8 | 99.8 | |
| LAVEModel backbone=DiffuC-7B2026.01 | 99.6 | 100 | 100 | 100 | |
| LAVEModel backbone=LLaDA-8B2026.01 | 99.5 | 99.6 | 99.6 | 99.6 | |
| LAVEModel backbone=Dream-7B2026.01 | 99 | 99.3 | 99.6 | 99.7 | |
| FS-CDModel backbone=DiffuC-7B2026.01 | 96.9 | 97.3 | 98.1 | 98.1 | |
| FS-CDModel backbone=Dream-7B2026.01 | 96.8 | 97.9 | 98.2 | 98.2 | |
| FS-CDModel backbone=LLaDA-1.52026.01 | 96 | 97.2 | 97.7 | 98.1 | |
| FS-CDModel backbone=LLaDA-8B2026.01 | 94.9 | 95.7 | 96.4 | 96.4 | |
| IG-CDModel backbone=Dream-7B2026.01 | 94.9 | 95.3 | 95.5 | 95.6 | |
| IG-CDModel backbone=LLaDA-8B2026.01 | 94.8 | 95.1 | 95.7 | 96.7 | |
| IG-CDModel backbone=LLaDA-1.52026.01 | 94 | 95.1 | 95.4 | 96.2 | |
| IG-CDModel backbone=DiffuC-7B2026.01 | 92.1 | 92.5 | 92.7 | 92.7 | |
| NO-CDModel backbone=DiffuC-7B2026.01 | 89.9 | 90.2 | 90.4 | 90.4 | |
| NO-CDModel backbone=Dream-7B2026.01 | 86.8 | 88.9 | 89.3 | 89.3 | |
| NO-CDModel backbone=LLaDA-1.52026.01 | 86 | 87.1 | 87.7 | 87.9 | |
| NO-CDModel backbone=LLaDA-8B2026.01 | 85.4 | 88.9 | 90.1 | 96.4 |