Text Generation on ReedsyPrompts (test)
49.3RougeLUAG (global)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| UAG (global)Base Model=Llada-8B, variant=global2026.04 | 49.3 | 21.2 | 15.7 | |
| UAG (global)Base Model=Llama-3B, variant=global2026.04 | 31.2 | 44.2 | 38.8 | |
| NaiveBase Model=Llama-3B2026.04 | 25.7 | 29.3 | 0.5 | |
| UAG (local)Base Model=Llama-3B, variant=local2026.04 | 21.2 | 39.1 | 0 | |
| Top-pBase Model=Llama-3B2026.04 | 19.8 | 47.6 | 19.3 | |
| NaiveBase Model=Llada-8B2026.04 | 18 | 31.6 | 56.4 | |
| Avoidance DecodingBase Model=Llama-3B2026.04 | 12.65 | 82 | 25 | |
| UAGBase Model=Llama-3B2026.04 | 9.3 | 68 | 56.5 | |
| UAG (local)Base Model=Llada-8B, variant=local2026.04 | 8.2 | 27.8 | 10 | |
| UAGBase Model=Llada-8B2026.04 | 6.7 | 43.4 | 32.3 | |
| Top-kBase Model=Llama-3B2026.04 | 0.6 | 93 | 98 |