Grounded Text Generation on RAGTruth
33.14F1 ScoreToken-Guard
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Token-GuardBackbone=13B2026.01 | 33.14 | 30.41 | |
| Tree-of-ThoughtBackbone=13B2026.01 | 32.93 | 23.7 | |
| Chain-of-ThoughtsBackbone=3B2026.01 | 32.7 | 25.48 | |
| Chain-of-ThoughtsBackbone=13B2026.01 | 32.25 | 27.67 | |
| Guided DecodingBackbone=3B2026.01 | 30.5 | 18.7 | |
| Guided DecodingBackbone=13B2026.01 | 29.34 | 22.01 | |
| Tree-of-ThoughtBackbone=3B2026.01 | 27.05 | 21.16 | |
| BaseModelBackbone=13B2026.01 | 22.47 | 15.26 | |
| Predictive DecodingBackbone=3B2026.01 | 21.72 | 14.59 | |
| Token-GuardBackbone=3B2026.01 | 15.66 | 13.45 | |
| BaseModelBackbone=3B2026.01 | 12.11 | 6.29 |