Abstract
Large language models face challenges in maintaining accuracy and relevance due to the emergence of new information and the limitations of offline training data. Existing methods, such as reinforcement learning from human feedback or standard retrieval augmented generation, often lack real-time correction capabilities or rely on static data sources.
A method is disclosed for real-time model refinement using a combined machine learning and human learning architecture. Vetted domain experts are integrated into a live feedback loop to provide grounded data for model outputs. A consensus-based review system is utilized to validate corrections through a required number of expert approvals. Contributions are managed via an impact tracking engine that monitors measurable improvements in output reliability. Integration is achieved by performing similarity searches on a curated knowledge base and prepending verified snippets to model prompts. This architecture ensures that model responses remain accurate and aligned with current developments through continuous, expert-validated updates.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Priya, Saloni, "Real-Time Large Language Model Refinement via Machine Learning and Human Learning Architecture", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/9935