Question Answering Retrieval on STaRK human-generated
55.8Hit@1AF-Retriever
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| AF-RetrieverLLM Backbone=GPT OSS 120B, Embedding=text-embedding-3-small, Protocol=Zero-shot, lmax=100, k=202025.05 | 55.8 | 66.4 | 60.7 | |
| 4StepFocusData Version=10%-version2025.05 | 52.9 | 61.5 | 50.3 | |
| KARProtocol=Zero-shot / One-shot2025.05 | 52.6 | 63.9 | 57.6 | |
| AvaTaRProtocol=Zero-shot / One-shot2025.05 | 41.6 | 56.9 | 48.5 | |
| VSS + RerankerData Version=10%-version2025.05 | 39.9 | 55.5 | 46.4 | |
| ReActProtocol=Zero-shot / One-shot2025.05 | 31.6 | 48.4 | 40.3 | |
| ReflexionProtocol=Zero-shot / One-shot2025.05 | 31.5 | 45.5 | 39.8 | |
| VSS (Ada-002)Embedding=Ada-0022025.05 | 28.5 | 46.8 | 38.3 | |
| DPRProtocol=Zero-shot / One-shot2025.05 | 7.6 | 19.4 | 14.1 |