Multi-hop Question Answering on IIRC
29EMCheckRLM
Evaluation Results
| Method | Links | ||
|---|---|---|---|
| CheckRLMBackbone=QwQ-32B2026.07 | 29 | 33.1 | |
| Q-DREAMBackbone=ChatGPT2025.05 | 28.2 | 31.9 | |
| CheckRLMBackbone=Qwen3-32B2026.07 | 27.8 | 31.6 | |
| CheckRLMBackbone=Qwen3-8B2026.07 | 25.4 | 29.7 | |
| Search-o1Backbone=QwQ-32B2026.07 | 25 | 29.2 | |
| Q-DREAMBackbone=Llama2-7B2025.05 | 23.8 | 27.2 | |
| Vanilla RAGBackbone=QwQ-32B2026.07 | 22 | 25 | |
| ChatQA2Backbone=Llama3-8B2025.05 | 21.4 | 25.7 | |
| Direct ReasoningBackbone=QwQ-32B2026.07 | 21.4 | 24.8 | |
| Self-RAGBackbone=QwQ-32B2026.07 | 21.3 | 25.1 | |
| IRCoTBackbone=ChatGPT2025.05 | 21 | 31 | |
| SUREBackbone=ChatGPT2025.05 | 20.6 | 25.5 | |
| Direct ReasoningBackbone=Qwen3-32B2026.07 | 20.2 | 23 | |
| Vanilla RAGBackbone=Qwen3-32B2026.07 | 19.8 | 22.5 | |
| Direct ReasoningBackbone=Qwen3-8B2026.07 | 18.2 | 21 | |
| RATBackbone=QwQ-32B2026.07 | 17 | 21.7 | |
| Vanilla RAGBackbone=Qwen3-8B2026.07 | 16.8 | 19.8 | |
| ReActBackbone=QwQ-32B2026.07 | 16.2 | 21.9 | |
| InstructRAGBackbone=Llama3-8B2025.05 | 14.2 | 18.1 | |
| ChatGPTBackbone=GPT-3.5-turbo2025.05 | 11.4 | 13.8 | |
| FLAREBackbone=QwQ-32B2026.07 | 9.6 | 12.3 |