Question Answering on NQ (Cover Exact Match)
0.574Cover EMStructRAG
Evaluation Results
| Method | Links | |
|---|---|---|
| StructRAGBackbone=Llama3.1-70B-Instruct2026.01 | 0.574 | |
| DeepNoteBackbone=Llama3.1-70B-Instruct2026.01 | 0.566 | |
| Iter-RetGenBackbone=Qwen3-32B2026.01 | 0.565 | |
| PAGERBackbone=Qwen3-32B2026.01 | 0.565 | |
| PAGERBackbone=Llama3.1-70B-Instruct2026.01 | 0.564 | |
| Vanilla RAGBackbone=Llama3.1-70B-Instruct2026.01 | 0.557 | |
| DeepNoteBackbone=Qwen3-32B2026.01 | 0.556 | |
| StructRAGBackbone=Qwen3-32B2026.01 | 0.551 | |
| Vanilla RAGBackbone=Qwen3-32B2026.01 | 0.546 | |
| RATBackbone=Llama3.1-70B-Instruct2026.01 | 0.536 | |
| RATBackbone=Qwen3-32B2026.01 | 0.532 | |
| Iter-RetGenBackbone=Llama3.1-70B-Instruct2026.01 | 0.53 | |
| IRCoTBackbone=Qwen3-32B2026.01 | 0.522 | |
| Search-o1Backbone=Llama3.1-70B-Instruct2026.01 | 0.516 | |
| Vanilla LLMBackbone=Llama3.1-70B-Instruct2026.01 | 0.503 | |
| IRCoTBackbone=Llama3.1-70B-Instruct2026.01 | 0.499 | |
| Search-o1Backbone=Qwen3-32B2026.01 | 0.492 | |
| Vanilla LLMBackbone=Qwen3-32B2026.01 | 0.362 |