Automated Essay Feedback System in Cross Prompt Settings

An AI driven essay feedback system should provide well-organized and informative feedback with suggestions in areas to improve. To achieve the desired results, we finetuned ChatGPT3.5-turbo on our Feedback dataset, tailoring it specifically for feedback generation. The paper discusses Chain-of-Thought (CoT) and Chunk-based prompting techniques for generating rubric based and holistic feedback. While Chunkbased analysis provides rubric-specific feedback more effectively, CoT based analysis provides better holistic feedback.. These techniques enable the model to evaluate various linguistic features while providing meaningful, context-driven suggestions. This work will emphasize the advantages and disadvantages of both methods and their usability for essay evaluation platforms. These findings can be applied in adaptive learning systems as well as in platforms helping to grade using AI. We suggest combining these methods to provide more effective and detailed feedback in real-life applications.