AI-based Supreme Court Appeal Predictor with Retrieval-Augmented Explainability: Advancing Beyond ILDC for Practical Legal Decision Support

Automated legal judgment prediction systems have received considerable research interest; yet, their applicability is limited due to poor explainability, ungrounded legislation, and unsatisfactory real-world integration. Although the ILDC dataset proposed the Court Judgment Prediction and Explanation (CJPE) task, current methods are primarily concerned with offline evaluation instead of practical judicial support systems. This paper proposes an AI-Based Supreme Court Appeal Predictor that expands the CJPE challenge with retrieval-augmented reasoning, statutory mapping, and judicial explanation structuring. The proposed system combines a fine-tuned InLegalBERT-based transformer classifier, FLAN-T5-based legal reasoning component, BNS statutory inference engine, and FAISS-enabled precedent retrieval. Differing from previous methods, our system generates court like structured judgments, locates relevant statutory sections, retrieves similar precedents, and generates human-readable legal reasoning amenable to legal processes. Experimental evaluation shows that our proposed system achieves substantial improvements in practical explainability and legal usability over baseline CJPE methods. This research brings the legal AI community one step closer to real-world judicial support systems instead of simple text classification tasks.