Research map / Large language models / Retrieval and knowledge
MIRAGE: Scaling Test-Time Inference with Parallel Graph-Retrieval-Augmented Reasoning Chains
Kaiwen Wei, Rui Shan, Dongsheng Zou, Jianzhong Yang, Bi Zhao, Junnan Zhu, Jiang Zhong
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| graph-based multi-chain retrieval-augmented reasoning (MIRAGE) vs Search-o1 | QwQ-32B · ExplainCPE | accuracy (%): 80.7% -> 84.8% | Improved |
| MIRAGE vs Search-o1 | QwQ-32B · GenMedGPT-5k | GPT-4o rank: 3.3 -> 1.8 | Improved |
| MIRAGE vs MindMap | QwQ-32B · ExplainCPE | accuracy (%): 84.6% -> 84.8% | No clear effect |
| MIRAGE vs MindMap on ExplainCPE rank | QwQ-32B · ExplainCPE | GPT-4o rank: 3.1 vs 3.1 | No clear effect |
| swap backbone to DeepSeek-R1-32B | MIRAGE · ExplainCPE | accuracy (%): 82.2% -> 84.4% | Improved |
| add Question Decomposer (ablation) | MIRAGE · GenMedGPT-5k pairwise | win rate vs ablated: win 40.72 vs lose 14.31 | Improved |
| add Answer Synthesizer with cross-chain verification (ablation) | MIRAGE · GenMedGPT-5k pairwise | win rate vs ablated: win 44.03 vs lose 12.73 | Improved |
| increase max sub-questions Nq (ablation) | MIRAGE · three medical QA benchmarks | accuracy and GPT-4o rank | Mixed |
| increase retrieval limit Nr (ablation) | MIRAGE · three medical QA benchmarks | accuracy and GPT-4o rank | Improved |
| QwQ-32B vs GPT-4o+ToT | QwQ-32B · ExplainCPE | accuracy (%): 80.2% -> 82.8% | Improved |
| add static BM25/dense retrieval to QwQ-32B | QwQ-32B · ExplainCPE | accuracy (%): 82.8% -> 83.5% (BM25), 82.4% (dense) | Mixed |
| Search-o1 web retrieval vs QwQ-32B | QwQ-32B · ExplainCPE | accuracy (%): 82.8% -> 80.7% | Hurt |
Closest papers by meaning
- Simple is Effective: The Roles of Graphs and Large Language Models in Knowledge-Graph-Based Retrieval-Augmented Generation (ICLR 2025 · similarity 0.87)
- Chiron-o1: Igniting Multimodal Large Language Models towards Generalizable Medical Reasoning via Mentor-Intern Collaborative Search (NeurIPS 2025 · similarity 0.87)
- Think-on-Graph 2.0: Deep and Faithful Large Language Model Reasoning with Knowledge-guided Retrieval Augmented Generation (ICLR 2025 · similarity 0.86)
- Retriever-Free Retrieval-Augmented Reasoning via Corpus-Traversing MCTS (NeurIPS 2026 · similarity 0.86)
- HyPER: Bridging Exploration and Exploitation for Scalable LLM Reasoning with Hypothesis Path Expansion and Reduction (ICML 2026 · similarity 0.86)
- Reasoning-Enhanced Healthcare Predictions with Knowledge Graph Community Retrieval (ICLR 2025 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
