Discourse Question Answering on NarrativeQA (F1 score)
26.6F1 ScoreSemFlowRAG
Evaluation Results
| Method | Links | |
|---|---|---|
| SemFlowRAGKnowledge Representation=Structured2026.06 | 26.6 | |
| HippoRAG 2 (reproduced)Knowledge Representation=Structured, Reproduction Status=reproduced2026.06 | 26.2 | |
| HippoRAG 2Knowledge Representation=Structured2026.06 | 25.9 | |
| NV-Embed-v2 (7B)Knowledge Representation=Embedding-based2026.06 | 25.7 | |
| GritLM-7BKnowledge Representation=Embedding-based2026.06 | 23.9 | |
| GraphRAGKnowledge Representation=Structured2026.06 | 23 | |
| RAPTORKnowledge Representation=Structured2026.06 | 21.4 | |
| GTE-Qwen2-7B-InstructKnowledge Representation=Embedding-based2026.06 | 21.3 | |
| GTRKnowledge Representation=Unstructured2026.06 | 19.9 | |
| ContrieverKnowledge Representation=Unstructured2026.06 | 19.7 | |
| BM25Knowledge Representation=Unstructured2026.06 | 18.3 | |
| HippoRAGKnowledge Representation=Structured2026.06 | 16.3 | |
| Llama-3.3-70B-Instruct w/o RetrievalKnowledge Representation=Unstructured, Retrieval=w/o Retrieval2026.06 | 12.9 | |
| LightRAGKnowledge Representation=Structured2026.06 | 3.7 |