Multi-hop Question Answering on Bamboogle out-of-domain (test)
68.8Accuracy (R)DLLM-Searcher
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| DLLM-SearcherCategory=dLLMs Agent, paradigm=P-ReAct, training=VRPO, backbone=SDAR2026.02 | 68.8 | 69.6 | |
| R1SearcherCategory=ARM-based LLMs Agent, setup=modified (5.1.3), search_mode=best of local/Google search2026.02 | 66.4 | 68.8 | |
| WebSailorCategory=ARM-based LLMs Agent, setup=modified (5.1.3), tools=search only2026.02 | 57.6 | 65.6 | |
| ReARTeRCategory=Traditional RAG2026.02 | 49.6 | 54.4 | |
| Search-o1Category=ARM-based LLMs Agent2026.02 | 49.6 | 52 | |
| Search-R1Category=ARM-based LLMs Agent2026.02 | 47.2 | 56 | |
| LLaDACategory=dLLMs Agent, paradigm=ReAct, acceleration=Fast-dLLM2026.02 | 46.4 | 42.4 | |
| CR-PlannerCategory=Traditional RAG2026.02 | 35.2 | 34.4 | |
| IRCoTCategory=Traditional RAG2026.02 | 32 | 39.2 | |
| Iter-RetGenCategory=Traditional RAG2026.02 | 26.4 | 32 | |
| Adaptive-RAGCategory=Traditional RAG2026.02 | 21.6 | 25 | |
| SuReCategory=Traditional RAG2026.02 | 17.6 | 28 | |
| Selective-ContextCategory=Traditional RAG2026.02 | 15.2 | 20.8 | |
| DreamCategory=dLLMs Agent, paradigm=ReAct, acceleration=Fast-dLLM2026.02 | 12 | 13.6 |