Multi-step Question Answering on HotpotQA (test)
66.7EMKeyExtract + CoopRAG
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| KeyExtract + CoopRAGReader (LLM)=GPT-4o-mini2025.12 | 66.7 | 79.2 | |
| KeyExtract + CoopRAGReader (LLM)=Gemma2-9B2025.12 | 64.9 | 77.9 | |
| IRCoT + CoopRAGReader (LLM)=GPT-4o-mini2025.12 | 64.3 | 75.5 | |
| IRCoT + HippoRAGReader (LLM)=GPT-3.5-turbo2025.12 | 45.7 | 59.2 | |
| IRCoT + ColBERTv2Reader (LLM)=GPT-4o-mini2025.12 | 45.5 | 58.4 |