Research map / Large language models / Retrieval and knowledge
ComoRAG: A Cognitive-Inspired Memory-Organized RAG for Stateful Long Narrative Reasoning
Juyuan Wang, Rongchen Zhao, Wei Wei, Yufeng Wang, Mo Yu, Jie Zhou, Jin Xu, Liyan Xu
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| add veridical layer to index (ablation) | ComoRAG · EN.MC | accuracy: 51.97 -> 72.93 | Improved |
| add semantic layer to index (ablation) | ComoRAG · EN.MC | accuracy: 64.63 -> 72.93 | Improved |
| add episodic layer to index (ablation) | ComoRAG · EN.MC | accuracy: 64.63 -> 72.93 | Improved |
| add metacognition (memory workspace) (ablation) | ComoRAG · EN.QA | F1: 26.95 -> 34.52 | Improved |
| add regulation (goal-oriented probing queries) (ablation) | ComoRAG · EN.MC | accuracy: 55.02 -> 72.93 | Improved |
| add both metacognition and regulation (ablation) | ComoRAG · EN.MC | accuracy: 54.15 -> 72.93 | Improved |
| add ComoRAG loop to RAPTOR (ablation) | RAPTOR · EN.MC | accuracy: 57.21 -> 69.00 | Improved |
| add ComoRAG loop to HippoRAGv2 (ablation) | HippoRAGv2 · EN.MC | accuracy: 60.26 -> 68.56 | Improved |
| use GPT-4.1 as LLM agent instead of GPT-4o-mini (ablation) | ComoRAG · EN.MC | accuracy: 72.93 -> 78.17 | Improved |
| use Qwen3-32B as LLM agent (ablation) | ComoRAG · EN.MC | accuracy: 72.93 -> 74.24 | Improved |
| larger embedding model (NV-Embed-v2 7B) instead of BGE-M3 for naive RAG vs ComoRAG with BGE-M3 (ablation) | naive RAG · EN.MC | accuracy: 61.13 vs 72.93 | No clear effect |
| single-step retrieval only (step 0) versus cognitive loop (ablation) | ComoRAG · EN.MC | accuracy | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- RAPTOR — RAPTOR: Recursive Abstractive Processing for Tree-Organized Retrieval (ICLR 2024) · better on 3 of 3 shared results
- HippoRAGv2 — HippoRAG: Neurobiologically Inspired Long-Term Memory for Large Language Models (NeurIPS 2024) · better on 3 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- REMem: Reasoning with Episodic Memory in Language Agent (ICLR 2026 · similarity 0.87)
- OmniToM: Benchmarking Theory of Mind in LLMs via Explicit Belief Modeling (NeurIPS 2026 · similarity 0.87)
- Episodic Memories Generation and Evaluation Benchmark for Large Language Models (ICLR 2025 · similarity 0.86)
- StructRAG: Boosting Knowledge Intensive Reasoning of LLMs via Inference-time Hybrid Information Structurization (ICLR 2025 · similarity 0.86)
- Longer Context, Deeper Thinking: Uncovering the Role of Long-Context Ability in Reasoning (NeurIPS 2025 · similarity 0.85)
- MC-Search: Evaluating and Enhancing Multimodal Agentic Search with Structured Long Reasoning Chains (ICLR 2026 · similarity 0.85)
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.
