Open-Ended Question Answering on DisastQA
71Base CoverageMistral-7B
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| Mistral-7BModel=Mistral-7B, Configuration=Base (No RAG)2026.04 | 71 | — | — | |
| Qwen-3-8BModel=Qwen-3-8B, Configuration=Base (No RAG)2026.04 | 70.3 | — | — | |
| LLaMA-3-8BModel=LLaMA-3-8B, Configuration=Base (No RAG)2026.04 | 68.2 | — | — | |
| DeepSeek-V3Model=DeepSeek-V3, Configuration=Base (No RAG)2026.04 | 59.8 | — | — | |
| DisastRAGBackbone=DeepSeek-V3, Best Configuration=Vector Retrieval IR100-Rerank102026.04 | — | 70.3 | 10.5 | |
| DisastRAGBackbone=Qwen-3-8B, Best Configuration=Hybrid Retrieval IR200-Rerank152026.04 | — | 79 | 8.7 | |
| DisastRAGBackbone=LLaMA-3-8B, Best Configuration=Hybrid Retrieval IR200-Rerank52026.04 | — | 75.6 | 7.4 | |
| DisastRAGBackbone=Mistral-7B, Best Configuration=Hybrid Retrieval IR150-Rerank52026.04 | — | 78.2 | 7.2 |