Tech

Anthropic Launches 'Citations' Feature for AI Accuracy 

  • Reading time: 2 min
the new Anthropic Citations feature a new member in the Anthropic’s Claude family, allowing developers to ground generated responses

On Thursday, the new Anthropic Citations feature was released by the AI company, welcomes as a new member in the Anthropic’s Claude family, allowing developers to ground generated responses in passages from source documents, such as emails or reports.  

The goal of the feature available on Anthropic’s API and Google’s Vertex AI platform, is to improve transparency and accurate content through clear references for any information used in generated responses. 

Citations in AI Models 

The Citations feature helps developers upload source documents that Anthropic Claude API models use when generating information. Anthropic AI features cite the exact parts of the generated documents, which offers detailed references to support the information. 

Anthropic Citations are very useful in summarizing documents, Q&A, and Customer support, especially where users need to check the reliability of AI generated responses. By adding Citations, the number of hallucinations coming along with AI generated content will be reduced.  

Anthropic Citations are now available only for Claude 3.5 Sonnet and Claude 3.5 Haiku, and the company has not yet made plans to develop the feature in the future. The Citations feature is available via Anthropic’s API and through Google’s Vertex AI platform, which makes such integration relatively easy for developers to build over their AI applications. 

Anthropic API access on Citations feature comes with additional cost, depending on length of the source documents. Supposedly, a 100-page document would cost $0.30 with Claude 3.5 Sonnet or $0.08 with Claude 3.5 Haiku API, according to the standard API costs put by Anthropic.  

Citations Value for Developers 

The Citations feature will benefit developers in their work where accuracy is a foundation for the success of any project, it also provides users with direct references to source documents to fact-check claims and guarantees responses from AI come from credible sources. 

Citations in AI models also focus on AI systems’ transparency, which is gaining importance with the increased integration of AI technologies into business processes. AI is becoming more integral in decision making, and Anthropic hopes that Citations will be more trusted by businesses in applications that require verifiable data.  

Inside Telecom provides you with an extensive list of content covering all aspects of the tech industry. Keep an eye on our Tech sections to stay informed and up-to-date with our daily articles.

Tags: