Precision Is Not Faithfulness: Coverage-Aware Evaluation of Grounded Generation with a Complete Oracle
2026-06-08 • Computation and Language
Computation and Language
AI summaryⓘ
The authors study how current methods check if AI models tell the truth, focusing on whether claims made are supported (precision). They find these methods miss how much important information the model actually covers (recall), allowing models to score well by saying very little. Using detailed data from Formula 1 races and weather forecasts, they show that current models often miss over half the relevant facts, even when asked to be thorough. The authors create a new combined score including both precision and recall, validate it, and offer a new method that helps models be both accurate and comprehensive without needing extra references.
faithfulness metricsprecisionrecallFormula 1 telemetrygrounded generationcoveragebenchmarkingpromptingoracle domainverifier-guided generation
Authors
Juan S. Santillana
Abstract
Reference-free faithfulness metrics verify each atomic claim a model makes against ground truth, and are increasingly used to evaluate grounded generation. We show they share a blind spot: they measure only precision -- are the stated claims supported? -- and therefore reward abstention, since a model can score near-perfect faithfulness by saying almost nothing. We make this measurable using Formula 1 telemetry, a domain where strategic ground truth is derived deterministically and, crucially, completely: for each decision we know the full set of facts that mattered. This completeness -- absent in open-domain faithfulness benchmarks -- lets us measure recall (coverage of the relevant facts) exactly, alongside precision. On a multilingual (EN/ES/PT) benchmark of 7,253 decision instances spanning 150 races, the most precise frontier model covers under half of the relevant facts and ranks last by F1, so requiring coverage reorders the systems; the same effect reappears in a second complete-oracle domain (NOAA weather forecasts). A prompt ablation shows the low coverage is not an under-prompting artifact: explicitly asking models to be thorough does not close the gap. We pair faithfulness with coverage into a single score, validate the metric (controlled perturbation; agreement across a model-free regex extractor and a cross-family LLM extractor, system-level Spearman 1.0), and give a verifier-guided generation method that improves precision and recall without references. We release the benchmark, structured annotations, metric, baselines, and an interactive demo.