Perplexity Research, in collaboration with turbopuffer, has introduced the pplx-embed-v2-context-9b-preview, an innovative contextual embedding model aimed at improving retrieval-augmented generation (RAG) processes. This model is unique in that it embeds each chunk of information with the entire document context, allowing for a more comprehensive retrieval of answers along with the necessary supporting evidence.
The key advancement lies in the model’s training signal, which shifts the focus from retrieving a single ‘gold passage’ to acquiring both the answer and its context. This approach is expected to significantly enhance the reliability of information retrieval in AI applications, making it easier for users to verify the accuracy of retrieved data.
Additionally, the model’s design ensures that it is deployable in various scenarios, offering flexibility for developers looking to integrate sophisticated AI capabilities into their systems. As the demand for accurate and contextually enriched information continues to grow, this release positions Perplexity as a notable player in the AI landscape.
Compiled automatically by the Tech AI Newsdesk from public AI-news sources and summarised in our own words.
