Multimodal Question Answering on Environmental Report
80.8ComprehensivenessMegaRAG
Evaluation results
| Method | Links | ||||
|---|---|---|---|---|---|
| MegaRAGComparison Baseline=LightRAG2025.11 | 80.8 | 92 | 79.2 | 90.4 | |
| MegaRAGComparison Baseline=NaiveRAG2025.11 | 72.8 | 92 | 64 | 80 | |
| MegaRAGComparison Baseline=GraphRAG2025.11 | 68.8 | 81.6 | 93.6 | 84.8 | |
| MegaRAGComparison Baseline=GME2025.11 | 48.8 | 75.2 | 50.4 | 64 | |
| MegaRAGComparison Baseline=ColQwen2025.11 | 37.6 | 57.6 | 44.8 | 50.4 | |
| MegaRAGComparison Baseline=VisRAG2025.11 | 27.2 | 60.8 | 32 | 40 | |
| MegaRAGevaluation_protocol=Pairwise LLM-judge win rates, format=Query-driven Win / MegaRAG Win / Tie2025.11 | 6.7 | 12.4 | 14.7 | 5.5 | |
| Query-driven MMKGevaluation_protocol=Pairwise LLM-judge win rates, format=Query-driven Win / MegaRAG Win / Tie2025.11 | 6.7 | 12.4 | 14.7 | 5.5 | |
| VisRAGComparison Baseline=MegaRAG2025.11 | 4 | 15.2 | 32.8 | 9.6 | |
| GMEComparison Baseline=MegaRAG2025.11 | 2.4 | 12 | 24 | 5.6 | |
| ColQwenComparison Baseline=MegaRAG2025.11 | 2.4 | 17.6 | 25.6 | 9.6 | |
| GraphRAGComparison Baseline=MegaRAG2025.11 | 0.8 | 7.2 | 1.6 | 0.8 | |
| NaiveRAGComparison Baseline=MegaRAG2025.11 | 0 | 2.4 | 12.8 | 1.6 | |
| LightRAGComparison Baseline=MegaRAG2025.11 | 0 | 1.6 | 4.8 | 0 |