RAG Reranking on HotpotQA
69.74PrecisionCross-Encoder Reranker
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Cross-Encoder RerankerGeneration LLM=gpt-4o-mini2026.06 | 69.74 | 69.84 | 67.57 | 72.5 | |
| LLM-based RerankerGeneration LLM=gpt-4o-mini2026.06 | 69.46 | 69.29 | 67.2 | 71.9 | |
| No RerankerGeneration LLM=gpt-4o-mini2026.06 | 67.49 | 67.31 | 65.22 | 69.4 | |
| Ricci-FiltrationGeneration LLM=gpt-4o-mini2026.06 | 65.81 | 63.97 | 63.36 | 67 | |
| Cross-Encoder RerankerGeneration LLM=Llama 3.1-8B Instruct2026.06 | 62.55 | 59.57 | 58.98 | 64 | |
| LLM-based RerankerGeneration LLM=Llama 3.1-8B Instruct2026.06 | 62.45 | 60.63 | 59.49 | 64.4 | |
| Ricci-FiltrationGeneration LLM=Llama 3.1-8B Instruct2026.06 | 62.39 | 59.21 | 59.13 | 63 | |
| No RerankerGeneration LLM=Llama 3.1-8B Instruct2026.06 | 59.43 | 57.97 | 56.54 | 61.5 |