Long-form QA on Long-form QA Short Q, Long A (test)
6.182GPT4 ScoreWorldCup
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| WorldCupBackbone=GEMMA2-9B-IT, k=12026.02 | 6.182 | 83.1 | 76 | |
| WorldCupBackbone=GEMMA2-9B-IT, k=22026.02 | 6.136 | 88.6 | 80 | |
| GEMMA2-9B-ITWatermark=None2026.02 | 6.027 | — | — | |
| BiMarkBackbone=GEMMA2-9B-IT2026.02 | 5.946 | 88.3 | 76.2 | |
| MPACBackbone=GEMMA2-9B-IT2026.02 | 5.92 | 83.8 | 77 | |
| WorldCupBackbone=MINISTRAL-8B-IT, k=12026.02 | 5.319 | 97.7 | 87.1 | |
| WorldCupBackbone=LLAMA3.1-8B-IT, k=12026.02 | 5.258 | 95.2 | 84.3 | |
| MINISTRAL-8B-ITWatermark=None2026.02 | 5.25 | — | — | |
| LLAMA3.1-8B-ITWatermark=None2026.02 | 5.243 | — | — | |
| WorldCupBackbone=MINISTRAL-8B-IT, k=22026.02 | 5.207 | 99.6 | 90.5 | |
| WorldCupBackbone=LLAMA3.1-8B-IT, k=22026.02 | 5.149 | 98.6 | 88.2 | |
| MPACBackbone=MINISTRAL-8B-IT2026.02 | 5.122 | 94.6 | 87.9 | |
| MPACBackbone=LLAMA3.1-8B-IT2026.02 | 5.11 | 94 | 85.5 | |
| BiMarkBackbone=LLAMA3.1-8B-IT2026.02 | 5.066 | 98.1 | 87.2 | |
| BiMarkBackbone=MINISTRAL-8B-IT2026.02 | 4.998 | 99.2 | 89.4 |