Tag
Build2SPARQL introduces a large-scale KG-grounded benchmark (6,136 validated SPARQL queries, 30,680 questions) for text-to-SPARQL querying over building knowledge graphs, where SPARQL queries are generated and validated by graph-traversal code and LLMs only produce natural-language questions. Retrieval-augmented evaluation of three open-weight models lifts exact-match accuracy from 0.2–20% zero-shot to 56–65% with three-shot retrieved context.
This paper presents a self-improving 'researcher agent' for Text-to-SPARQL question answering over knowledge graphs, which iteratively refines its own prompts and tools. Evaluated on DBpedia, it achieves 0.22 accuracy and identifies predicate selection as the main bottleneck.
This paper proposes GGC, a Generator–Gate–Corrector framework that selectively corrects LLM-generated SPARQL queries to improve reliability and accuracy, achieving 98.33% query-level accuracy on MCQA while reducing inference overhead by 45%.
This paper introduces Kontrast, a framework for automatically detecting knowledge inconsistencies across Wikipedia text, tables, and Wikidata knowledge graphs using Text-to-SPARQL and LLM reasoning.