Precision-recall
Also written: PR curve
Measures how accurately an AI identifies relevant items and finds all actual relevant items, crucial for classification tasks.
Precision and recall are two ways to measure how good an AI is at finding specific things, like identifying spam emails or relevant search results.
Precision tells you how often the AI was right when it said it found something. If an AI flags 10 emails as spam and 9 of them actually are spam, its precision is 90%.
Recall tells you how many of the actual relevant things the AI managed to find. If there were 15 spam emails in total and the AI found 9 of them, its recall is 60%. Both numbers are important because you might want an AI that is very precise (doesn't make many mistakes) or one that has high recall (finds most of what it's looking for).
Precision and recall are key metrics for evaluating classification models, particularly in imbalanced datasets. Precision measures the proportion of true positive predictions among all positive predictions made by the model (TP / (TP + FP)). Recall, also known as sensitivity or true positive rate, measures the proportion of actual positive instances that were correctly identified by the model (TP / (TP + FN)).
There is often a trade-off between precision and recall. Increasing one may decrease the other. The choice of which metric to prioritize depends on the specific application. For instance, in medical diagnosis, high recall is crucial to avoid missing any potential cases, even if it means more false positives. In contrast, for a spam filter, high precision might be preferred to avoid marking legitimate emails as spam.
Precision-recall metrics are crucial for evaluating AI systems that perform classification tasks, such as identifying bugs or categorizing information. As frontier AI labs develop more sophisticated models for these applications, objective evaluation becomes paramount. Benchmarks like the one from Microsoft are increasingly using precision-recall to demonstrate the effectiveness of their AI solutions.
- ReviewBench: An open benchmark for AI code reviewMicrosoft · GitHub Blog: Copilot ·
- Microsoft Copilot
- Microsoft Copilot and other AI development at Microsoft likely use precision-recall to measure the performance of models involved in tasks like code analysis or content moderation.
- Accuracy
- Accuracy measures the overall correctness of the model across all clas
- F1 Score
- The F1 score is the harmonic mean of precision and recall, providing a
- ROC Curve
- The ROC curve plots the true positive rate against the false positive
Precision and recall originated in the field of information retrieval and were formally defined by Alan Cooper in 1950. They became standard metrics for evaluating search engines and later classification algorithms in machine learning. Source ↗
Ask an AI assistant like ChatGPT or Copilot to generate a list of potential solutions for a specific coding problem. Then, manually check how many of the suggested solutions are actually correct (precision) and how many correct solutions were provided out of all possible correct solutions (recall).
The Playbook: reusable how-tos, prompts, configs and tips pulled out of the news.
- HeuristicChoosing an AI Code Reviewer Based on Workflow NeedsSelect an AI code reviewer by understanding its strengths in catching different types of issues and its precision-recall trade-offs. · 5 min
- PromptRole-play a student struggling with math for AI tutor simulationSimulate a student's math struggles and uncertainty for AI tutor evaluation. · 5 min
- TemplateReusable workflow templates for healthcare teamsTeams can create and share templates for common workflows like documentation or prior authorizations. · 15 min
- TemplateTemplate for building an AI tool tailored to your workA structured approach to defining and building a custom AI tool for specific business needs. · 15 min
- TemplateEvals CSV prompt list for agent skillsA CSV file template for listing prompts used in evaluating agent skills. · 20 min
Questions people ask
- What is the difference between precision and recall?
- Precision measures how many of the AI's positive predictions were correct. Recall measures how many of the actual positive cases the AI found.
- Why are both precision and recall important?
- An AI could be very precise but miss many actual cases (low recall), or find most cases but include many incorrect ones (low precision). Both metrics together give a fuller picture of performance.
First seen on Subvolts Oct 5, 2026. Updated Oct 12, 2026. Written by an AI model from the items above and reviewed against them weekly; report an error.