By now, everyone knows that single-vector embedding models are hugely limiting for modern workflows.
Mixedbread
Visit on X@mixedbreadai4.1K followers
41,133 views
By now, everyone knows that single-vector embedding models are hugely limiting for modern workflows.
But they contain than you think: you can extract sparse Latent Terms from them.
And it turns out that BM25 is all you need to turn this vocabulary into a strong retriever.
Comments
Comments from the original post on X (Twitter)
No comments yet
