A controlled prompting protocol shows open-weight LLMs rely heavily on memorized Wikidata entities when writing SPARQL queries, with accuracy collapsing on a rarely used benchmark.
Available at SSRN 5009450
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.IR 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
SPARQL Query Generation with LLMs: Measuring the Impact of Training Data Memorization and Knowledge Injection
A controlled prompting protocol shows open-weight LLMs rely heavily on memorized Wikidata entities when writing SPARQL queries, with accuracy collapsing on a rarely used benchmark.