Fact Verification on TruthfulQA
82.7Macro F1SEVA-GRPO (3B)
Evaluation Results
| Method | Links | |
|---|---|---|
| SEVA-GRPO (3B)Output format=struct, Size=3B2026.06 | 82.7 | |
| SEVA-SFT (3B)Output format=struct, Size=3B2026.06 | 72.1 | |
| MiniCheck-FT5Output format=binary2026.06 | 59.5 | |
| GPT-4o-miniOutput format=struct2026.06 | 48.6 |