Anthropic Launches Citations Feature to Enhance AI Accuracy

The new API feature allows Claude models to ground responses in source documents, aiming to reduce errors and improve trustworthiness.

Anthropic has introduced a Citations feature for its Claude 3.5 Sonnet and Haiku models, available through its API and Google Cloud's Vertex AI platform.
The feature enables AI models to cite specific sentences and passages from source documents, improving the reliability of generated responses.
Early adopters, including Thomson Reuters and financial technology company Endex, report promising results in reducing hallucinations and increasing citation accuracy.
Citations uses Anthropic's token-based pricing model, but quoted text in responses does not count toward output token costs.
The feature is designed to assist developers in building tools for document summarization, complex queries, and customer support systems.