Multi-step Question Answering on MuSiQue (test)
53.8Exact MatchKeyExtract + CoopRAG
Evaluation results
| Method | Links | ||
|---|---|---|---|
| KeyExtract + CoopRAGReader (LLM)=GPT-4o-mini2025.12 | 53.8 | 68.6 | |
| KeyExtract + CoopRAGReader (LLM)=Gemma2-9B2025.12 | 52.5 | 66.1 | |
| IRCoT + CoopRAGReader (LLM)=GPT-4o-mini2025.12 | 52.3 | 65.9 | |
| IRCoT + HippoRAGReader (LLM)=GPT-3.5-turbo2025.12 | 21.9 | 33.3 | |
| IRCoT + ColBERTv2Reader (LLM)=GPT-4o-mini2025.12 | 19.1 | 30.5 |