Breakthrough solution enables enterprises to control how AI systems prioritize retrieved
information with unprecedented precision
Contextual AI, the Enterprise AI company, announced the launch of the first instruction-following reranker, setting a new standard for retrieval accuracy in enterprise RAG (retrieval-augmented generation) systems.
Enterprise AI now faces significant challenges when dealing with conflicting information from messy, inconsistent, and constantly evolving data sources. Traditional ranking methods are designed to maximize retrieval relevancy, but fail to do so effectively because they lack the ability to follow specified preferences, often resulting in suboptimal outputs.
“In critical enterprise environments, prioritizing the wrong information can lead to severe consequences,” said Douwe Kiela, CEO of Contextual AI. “Our instruction-following reranker puts precise control in the hands of users, allowing them to guide AI systems with natural language instructions that reflect their specific business needs and domain expertise.”
Also Read: Observe Debuts Frontend Observability for UX & Debugging
The groundbreaking solution enables users to provide natural language instructions that control how retrieved information is ranked, such as “Prioritize recent documents,” “Prefer PDFs over other sources,” or “Give more weight to documents from our CEO.” This level of control is crucial for enterprises operating in domains where accuracy is paramount, such as financial services—where it can prevent erroneous trade recommendations—or healthcare—where it ensures the most relevant patient information is used for treatment planning.
“At ClaimWise, we work with complex patents and scientific articles. Meeting the quality expectations of our users has always been one of our main technical challenges,” said Milo Moerkerke, Co-founder at Claimwise. “The Contextual AI reranker’s instruction-following capability is great; we have some nuanced requirements about how to prioritize different types of paragraphs for different fields of invention, and the reranker follows those instructions well.”
Beyond its innovative instruction-following capabilities, Contextual AI’s reranker has achieved state-of-the-art results on the industry-standard BEIR benchmark, outperforming all other rerankers currently on the market.
The instruction-following reranker is available immediately as an API on the Contextual AI Platform. The solution integrates seamlessly with existing enterprise AI systems, requiring minimal implementation time while delivering immediate improvements to information retrieval quality.
Source: PRNewswire
