Copyright © 2026 Authors retain the copyright of this article. This article is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

@article{200357,
        author = {Pranav Patil and Rony Preetam and Amit Patil and G. Banidhar and Ketan Bemalkhedkar},
        title = {Automated Knowledge Synthesis and Collaborative Discovery: A Unified Study of STORM and Co-STORM Frameworks for Grounded Long-Form Report Generation},
        journal = {International Journal of Innovative Research in Technology},
        year = {2026},
        volume = {12},
        number = {12},
        pages = {1119-1127},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=200357},
        abstract = {Producing structured, factually grounded long-form articles entirely from scratch continues to pose significant difficulties for contemporary large language model (LLM) pipelines, chiefly because the cognitively demanding pre-writing stage—source identification, perspective mapping, question formulation, and evidence-based outline construction—remains poorly automated. This paper offers a unified examination of two interrelated systems that tackle complementary aspects of the knowledge-synthesis problem. STORM (Synthesis of Topic Outlines through Retrieval and Multi-perspective Question Asking) mechanises the pre-writing phase by orchestrating simulated multi-perspective dialogues between persona-driven writers and a retrieval-grounded expert, subsequently assembling a hierarchical outline before article drafting begins. Co-STORM extends this paradigm into a collaborative, human-steerable discourse model: users observe and intermittently redirect conversations among several LM agents while a dynamic mind map tracks the evolving knowledge landscape in real time. Both frameworks leverage the DSPy prompting toolkit and retrieval-augmented generation (RAG). Evaluation on the FreshWiki benchmark and the newly curated WildSeek dataset confirms that STORM exceeds outline-driven RAG baselines in breadth and organisational coherence, and that Co-STORM surpasses conventional search engines and RAG chatbots in depth, novelty, and user-reported satisfaction. Expert Wikipedia editors and lay user cohorts alike express strong preference for the outputs of both systems, substantiating the practical viability of LLM-driven knowledge-synthesis pipelines for scholarly and encyclopaedic writing tasks.},
        keywords = {Large Language Models, Retrieval-Augmented Generation, Long-Form Text Generation, Collaborative Information Seeking, Pre-Writing Automation, Wikipedia-Like Article Generation, Multi-Agent Systems, Knowledge Discovery.},
        month = {May},
        }