Paper Title
Multi-Block Retrieval-Augmented Generation for Adaptive Learning Path Synthesis: A Phase-Wise Approach to Personalized Education

Abstract
This paper presents a novel multi-block Retrieval-Augmented Generation (RAG) framework for generating adaptive, personalized learning pathways for technical education. Unlike conventional static recommendation systems or single-stage generation approaches, our method employs specialized RAG blocks for targeted information extraction, iterative phase-wise roadmap construction, and multi-dimensional user profiling. The system integrates ChromaDB vector storage with Large Language Models (LLMs) to dynamically retrieve contextual information across multiple knowledge domains. We introduce a hierarchical generation pipeline that decomposes learning goals into five sequential phases, each informed by vector-retrieved context and previous phase outcomes. Experimental evaluation involving a user study (N=30) and automated semantic analysis indicates significantly increased goal clarity and resource relevance compared to static baselines. Furthermore, the specialized RAG architecture achieved a 4.1x increase in task granularity and a 0.89 Semantic Coherence score (cosine similarity), thus outperforming single-stage LLM generation and rule-based systems. Keywords - Retrieval-Augmented Generation, Personalized Learning, Adaptive Education Systems, Vector Databases, Learning Path Optimization, Educational Technology