Lexara-RF: Reference-Free Metrics for Evaluating Conversational Visual Analytics Agents
3.40T1 sourcearXiv cs.HC
Source record
Published by arXiv cs.HC (T1 source). The original is at https://arxiv.org/abs/2609.17842.
Pipeline notes
The summary and note below are generated by the signal pipeline — they are Beyond Desk’s reading, not quotations from the source.
SummaryLexara-RF introduces 13 reference-free metrics for evaluating conversational visual analytics (CVA) agents, reformulating evaluation as verification via consistency, intent-alignment, and design-validity checks grounded in visualization theory and Gricean principles. Validated on a human-rated corpus, it matches reference-based formulations and outperforms NLG surface-similarity baselines.
Why it mattersOffers a practical evaluation alternative to costly curated benchmarks for CVA agents, with empirically validated metrics and grounded failure localization.
Cited by
No citations on record.