Question Answering on AstroQA (Hard)
83.59AccuracyMistral-7B + AstroRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| Mistral-7B + AstroRAGBackbone=ViT, Zero-shot=true2026.05 | 83.59 | 83.5 | |
| AstroRAGBackbone=Mistral-7B2026.05 | 83.59 | 83.5 | |
| AstroSage-Llama-3.1-8B + AstroRAGBackbone=MPT, Zero-shot=true2026.05 | 81.3 | 81.2 | |
| DrQA2026.05 | 62.6 | 62.37 | |
| Map-Reduce2026.05 | 60.69 | 60.98 | |
| Llama 2-Chat + AstroRAGBackbone=MPT, Zero-shot=true2026.05 | 54.2 | 54.07 | |
| AstroSage-Llama-3.1-8BBackbone=MPT, Zero-shot=true2026.05 | 39.31 | 38.53 | |
| Llama 2-ChatBackbone=MPT, Zero-shot=true2026.05 | 27.1 | 21.01 | |
| Mistral-7BBackbone=ViT, Zero-shot=true2026.05 | 24.05 | 10.26 | |
| BaselineBackbone=Mistral-7B2026.05 | 24.05 | 10.26 |