Question Answering on AmbigQA (Cover Exact Match)
60Cover EMPAGER
Evaluation Results
| Method | Links | |
|---|---|---|
| PAGERBackbone=Llama3.1-70B-Instruct2026.01 | 60 | |
| DeepNoteBackbone=Llama3.1-70B-Instruct2026.01 | 58.4 | |
| StructRAGBackbone=Llama3.1-70B-Instruct2026.01 | 58.2 | |
| Vanilla RAGBackbone=Llama3.1-70B-Instruct2026.01 | 57 | |
| PAGERBackbone=Qwen3-32B2026.01 | 56.4 | |
| RATBackbone=Llama3.1-70B-Instruct2026.01 | 56.4 | |
| DeepNoteBackbone=Qwen3-32B2026.01 | 55.9 | |
| RATBackbone=Qwen3-32B2026.01 | 55.5 | |
| Iter-RetGenBackbone=Qwen3-32B2026.01 | 55.4 | |
| Vanilla RAGBackbone=Qwen3-32B2026.01 | 54.8 | |
| StructRAGBackbone=Qwen3-32B2026.01 | 54.5 | |
| Search-o1Backbone=Llama3.1-70B-Instruct2026.01 | 53.8 | |
| Iter-RetGenBackbone=Llama3.1-70B-Instruct2026.01 | 52.3 | |
| IRCoTBackbone=Qwen3-32B2026.01 | 51.8 | |
| Search-o1Backbone=Qwen3-32B2026.01 | 51.5 | |
| IRCoTBackbone=Llama3.1-70B-Instruct2026.01 | 50.3 | |
| Vanilla LLMBackbone=Llama3.1-70B-Instruct2026.01 | 48.8 | |
| Vanilla LLMBackbone=Qwen3-32B2026.01 | 32.9 |