Grounded Text Generation on HaluEval
72.66F1 ScoreToken-Guard
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Token-GuardBackbone=13B2026.01 | 72.66 | 68.81 | |
| Token-GuardBackbone=3B2026.01 | 68.21 | 63.83 | |
| Guided DecodingBackbone=3B2026.01 | 63.47 | 59.11 | |
| Guided DecodingBackbone=13B2026.01 | 63.42 | 58.72 | |
| Tree-of-ThoughtBackbone=13B2026.01 | 62.2 | 58.07 | |
| Tree-of-ThoughtBackbone=3B2026.01 | 59.22 | 55.11 | |
| Predictive DecodingBackbone=3B2026.01 | 57.55 | 53.97 | |
| Chain-of-ThoughtsBackbone=3B2026.01 | 50.88 | 48.44 | |
| Chain-of-ThoughtsBackbone=13B2026.01 | 49.43 | 44.01 | |
| BaseModelBackbone=3B2026.01 | 42.95 | 39.95 | |
| BaseModelBackbone=13B2026.01 | 35.3 | 31.4 |