Precision
An evaluation metric for classification models that computes the ratio of correctly predicted positive instances to all instances predicted as positive. In IDP it can be used to monitor model hallucination (fabricated content).
Timeline (last 90 days)
For extraction tasks involving LLMs, when the model "fabricates" content, False Positives rise and Precision falls, so these metrics are more sensitive to hallucination monitoring, and the False Positive Rate should be watched to control hallucinations.
IDP scoring should not report only a single metric; at minimum it should give Precision, Recall, and F1 together and state the counting rules for blanks, NULLs, checkboxes, and unreadable values.
In RAG scenarios, Recall is usually prioritized over Precision, because missing a piece of key evidence often directly leads to an incorrect answer, whereas retrieving a few extra irrelevant documents only increases context noise and token cost.
All Facts (3)
For extraction tasks involving LLMs, when the model "fabricates" content, False Positives rise and Precision falls, so these metrics are more sensitive to hallucination monitoring, and the False Positive Rate should be watched to control hallucinations.
50%UnverifiedIDP scoring should not report only a single metric; at minimum it should give Precision, Recall, and F1 together and state the counting rules for blanks, NULLs, checkboxes, and unreadable values.
50%UnverifiedIn RAG scenarios, Recall is usually prioritized over Precision, because missing a piece of key evidence often directly leads to an incorrect answer, whereas retrieving a few extra irrelevant documents only increases context noise and token cost.
50%