Reducing Hallucinations in Language Model-based SPARQL Query Generation Using Post-Generation Memory Retrieval
Aditya Sharma, Luis Lara, Christopher J. Pal, Amal Zouaq · 2025
Read Reducing Hallucinations in Language Model-based SPARQL Query Generation Using Post-Generation Memory Retrieval by Aditya Sharma, Luis Lara, Christopher J. Pal, Amal Zouaq (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-13369.txt.
Description
Aditya Sharma and co-authors present PGMR, a 2025 framework that stops language models from inventing fake database addresses when writing queries. Turning a plain question into a SPARQL query over a knowledge graph is error-prone because models guess the exact identifiers from fuzzy memory.
Their fix splits the job. The model drafts the query with plain-language placeholders where identifiers belong. A separate memory module then looks up the true identifiers and fills them in, refusing when its confidence runs low. Query correctness rose across models and data sets, while invented identifiers nearly vanished. The memory stayed reliable even when padded with nine times as many distracting entries.
Catalog record
| CREATOR | Aditya Sharma, Luis Lara, Christopher J. Pal, Amal Zouaq |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 33 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2502-13369