Graph-based Question Answering on 23 Queries (Hold-out)
70Mean F1LLM Planner
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| LLM PlannerLLM Backbone=Claude Haiku 4.52026.06 | 70 | — | — | |
| Agentic RAGLLM Backbone=Claude Haiku 4.52026.06 | 66.2 | — | — | |
| TF-IDF RAG2026.06 | 38.8 | — | — | |
| Det. GraphRAGLLM Backbone=Claude Haiku 4.52026.06 | 37.9 | — | — |