Question Answering on MUSIQUE Unfiltered
47.62AccuracyBaseline
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| BaselineModel=Mistral 7B2026.05 | 47.62 | 49.86 | 2.48 | |
| BaselineModel=Qwen 2.5 32B2026.05 | 47.54 | 52.8 | 3.54 | |
| +ScramblerModel=Mistral 7B2026.05 | 47.37 | 49.78 | 2.48 | |
| +DistributedModel=Mistral 7B2026.05 | 47.29 | 49.6 | 2.48 | |
| +ScramblerModel=Qwen 2.5 32B2026.05 | 47.08 | 52.37 | 3.55 | |
| +DistributedModel=Qwen 2.5 32B2026.05 | 46.83 | 52.23 | 3.55 | |
| BaselineModel=Llama 3.1 8B2026.05 | 35.79 | 40.11 | 2.93 | |
| +DistributedModel=Llama 3.1 8B2026.05 | 35.79 | 40.15 | 2.93 | |
| +ScramblerModel=Llama 3.1 8B2026.05 | 35.79 | 40 | 2.93 | |
| +DistributedModel=Qwen 2 7B2026.05 | 30.49 | 35.7 | 15.73 | |
| BaselineModel=Qwen 2 7B2026.05 | 30.41 | 35.57 | 15.72 | |
| +ScramblerModel=Qwen 2 7B2026.05 | 30.33 | 35.6 | 15.91 |