ComoRAG: A Cognitive-Inspired Memory-Organized RAG for Stateful Long Narrative Reasoning

2025-08-14

AI Summary

Traditional RAG systems suffer from stateless, single-step retrieval, limiting their ability to model dynamic plot progression and evolving character relationships in long-form narratives. Method: We propose a state-aware iterative reasoning framework that introduces a dynamic working memory and a global memory pool, enabling cyclical updating through exploratory evidence acquisition and historical knowledge integration. Our approach incorporates LLM-driven exploratory query generation, multi-turn retrieval, and explicit runtime narrative state modeling. Contribution/Results: Evaluated on four long-narrative benchmarks exceeding 200K tokens, our method significantly outperforms strong baselines, achieving up to an 11% relative improvement—particularly excelling on complex questions requiring holistic narrative understanding. The framework endows RAG with human-like sustained reasoning capabilities.

Technology Category

• Data Mining & Knowledge Management: Conversational Systems for Recommendation & Retrieval
• Natural Language Processing: Question Answering
• Knowledge Representation and Reasoning: Common-Sense Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Narrative comprehension on long stories and novels has been a challenging domain attributed to their intricate plotlines and entangled, often evolving relations among characters and entities. Given the LLM's diminished reasoning over extended context and high computational cost, retrieval-based approaches remain a pivotal role in practice. However, traditional RAG methods can fall short due to their stateless, single-step retrieval process, which often overlooks the dynamic nature of capturing interconnected relations within long-range context. In this work, we propose ComoRAG, holding the principle that narrative reasoning is not a one-shot process, but a dynamic, evolving interplay between new evidence acquisition and past knowledge consolidation, analogous to human cognition when reasoning with memory-related signals in the brain. Specifically, when encountering a reasoning impasse, ComoRAG undergoes iterative reasoning cycles while interacting with a dynamic memory workspace. In each cycle, it generates probing queries to devise new exploratory paths, then integrates the retrieved evidence of new aspects into a global memory pool, thereby supporting the emergence of a coherent context for the query resolution. Across four challenging long-context narrative benchmarks (200K+ tokens), ComoRAG outperforms strong RAG baselines with consistent relative gains up to 11% compared to the strongest baseline. Further analysis reveals that ComoRAG is particularly advantageous for complex queries requiring global comprehension, offering a principled, cognitively motivated paradigm for retrieval-based long context comprehension towards stateful reasoning. Our code is publicly released at https://github.com/EternityJune25/ComoRAG

Problem

• Addresses narrative comprehension in long stories with complex plots
• Overcomes stateless retrieval limitations in traditional RAG methods
• Enables dynamic reasoning via iterative memory consolidation cycles

Innovation

• Dynamic memory workspace for iterative reasoning cycles
• Generates probing queries for exploratory paths
• Integrates retrieved evidence into global memory pool