TechGraphRAG: An Agentic Graph-Augmented RAG Framework for Technical Literature Reasoning
For researchers in specialized technical domains, this work provides a practical case study of agentic RAG that addresses evidence sufficiency and citation integrity, though it is an incremental application of existing techniques to a new domain.
The paper introduces TechGraphRAG, an agentic RAG framework for technical literature reasoning over a domain-specific corpus of ~2,100 papers. The system uses a 13-step autonomous pipeline with evidence scoring, external search, knowledge graph traversal, and self-correction to improve retrieval and generation quality.
This paper presents an agentic retrieval-augmented generation (RAG) framework for domain-specific technical reasoning support, instantiated over a curated corpus of approximately 2,100 academic papers in intelligent tires, vehicle dynamics, and vehicle control. Unlike conventional single-pass RAG systems, the proposed architecture employs a 13-step autonomous pipeline that classifies queries by intent, scores evidence sufficiency against a multi-dimensional rubric, performs agentic retry with drift-guarded query reformulation, searches external academic databases (Crossref, OpenAlex, Semantic Scholar) through iterative optimize--search--vet loops, traverses a Neo4j knowledge graph for relational context, verifies citation integrity, and applies post-generation quality checks with automatic regeneration. Key contributions include a 100-point evidence sufficiency scoring framework across five dimensions with relevance damping and hybrid rule-based/LLM review; a route-dependent external search architecture with iterative agentic loops; a knowledge graph constructed via LLM-based entity extraction and OpenAlex author validation with intra-corpus citation resolution; and a self-correcting generation loop with citation verification and quality assessment. The framework is presented as a practical, implemented case study illustrating how agentic, evidence-grounded RAG can support literature navigation and technical reasoning over large, domain-specific corpora.