Copyright © 2025 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{182546,
        author = {RATNESH KUMAR SHARMA and Prof Satya Singh},
        title = {ENHANCING HUMAN-AI COLLABORATION IN DOCUMENT CLASSIFICATION THROUGH RLHF AND RLAIF: A STUDY ON ADAPTIVE INTERACTION MODELS},
        journal = {International Journal of Innovative Research in Technology},
        year = {2025},
        volume = {12},
        number = {2},
        pages = {3418-3423},
        issn = {2349-6002},
        url = {https://ijirt.org/article?manuscript=182546},
        abstract = {In recent times, Reinforcement Learning from Human Feedback (RLHF) and Reinforcement Learning from AI Feedback (RLAIF) have emerged as potent frameworks to improve human-AI collaboration, especially in intricate tasks like document classification in healthcare and other critical fields. This paper examines a hybrid classification framework that combines RLHF and RLAIF inside a multi-model ensemble comprising Random Forest (RF), Support Vector Machine (SVM), Gradient Boosting (GB), and Naive Bayes (NB), with the objective of developing an adaptive and intelligent human-AI interaction model. The aim is to create a resilient document classification system that achieves high accuracy while conforming to human reasoning and expectations. The methodology entails training each classifier on pre-processed, human-labelled documents, utilising reinforcement learning to enhance model performance through iterative feedback from both humans and AI, and subsequently amalgamating the results via ensemble techniques. The assessment of a test set including 300 documents across 11 unique categories resulted in flawless performance metrics—100% precision, recall, and F1-score—indicating the absence of misclassifications. Nonetheless, critical reflection reveals a concern over potential overfitting attributed to the consistency of elevated scores, necessitating the incorporation of different and more complex validation datasets. The study implies that models with somewhat below perfect accuracy (e.g., 99%) are often more pragmatic and generalizable in real-world settings, especially in sensitive domains like healthcare. This study shows how RLHF and RLAIF can be used to create adaptive AI systems that improve document classification skills. Performance, interpretability, and generalisation must be balanced.},
        keywords = {AI Collaboration; Document Classification; Adaptive interaction models; RLHF; RLAIF},
        month = {July},
        }