Question Answering on PopQA (EM, F1, ROUGE)
41.6EMR1-Searcher
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| R1-SearcherTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 41.6 | 35.2 | 41.3 | |
| GRIPTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 38.6 | 37.5 | 38.4 | |
| GRIP w/o RLTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 38.1 | 37.1 | 37.6 | |
| RetRobustTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 34.1 | 35.1 | 40.4 | |
| InstructRAGTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 33.1 | 35.7 | 40.3 | |
| GPT-4oTraining status=Training-free2026.04 | 30.6 | 32.3 | 39.9 | |
| ETCTraining status=Training-free2026.04 | 30.5 | 32.5 | 37.5 | |
| GainRAGTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 30.1 | 33.3 | 38.1 | |
| RobustRAGTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 29.7 | 27.7 | 32.4 | |
| SFT-RAGTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 29.4 | 25.2 | 30.4 | |
| GPT-3.5 TurboTraining status=Training-free2026.04 | 29.1 | 31.1 | 35.3 | |
| Single RAGTraining status=Training-free2026.04 | 22.8 | 26.6 | 31.1 | |
| INFO-RAGTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 18.3 | 22.6 | 23 | |
| Self-RAGTraining status=Training-based, Backbone=LLaMA3-8B, Retriever=BM252026.04 | 18.1 | 22.3 | 22.8 | |
| InstructTraining status=Training-free2026.04 | 17.4 | 19.6 | 23.2 | |
| DRAGINTraining status=Training-free2026.04 | 15.5 | 16.8 | 19.8 | |
| FLARETraining status=Training-free2026.04 | 14.3 | 16 | 18.4 |