Designing for Imperfect Prompts: A Simulation-Based Evaluation of an LLM-Based Enterprise Data Assistant Using Knowledge Graphs
3.60T1 sourcearXiv cs.HC
Source record
Published by arXiv cs.HC (T1 source). The original is at https://arxiv.org/abs/2610.23304.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryDesign Science Research study evaluating an LLM-based enterprise data assistant deployed in a CFO unit's data lake. It compares schema-based and knowledge-graph-based semantic mediation for resolving imperfect user prompts, finding knowledge graph mediation outperforms on precision, recall, F1, and user intent preservation.
Why it mattersReal deployment evidence that knowledge graph mediation measurably improves how enterprise assistants handle the messy, domain-specific prompts business users actually type.
Cited by
No citations on record.