Improving Text Embeddings with Large Language Models

Interesting, but this aspect makes me double-take: &quot;We demonstrate that Mistral-7B, when fine-tuned solely on synthetic data, attains competitive performance on the BEIR [ 40 ] and MTEB [27] benchmarks&quot;.<p>E5&#x2F;BGE large are an order of magnitude smaller than Mistral-7B. So is this just &quot;bigger model wins&quot; in disguise?<p>I need to read the whole paper carefully, but this jumped out at me.

&gt; Subjects: Computation and Language (cs.CL); Information Retrieval (cs.IR)<p>I&#x27;m surprised they didn&#x27;t put `Machine Learning (cs.LG)` and `Machine Learning (stat.ML)`.

I am confused, aren&#x27;t LLMs already embeddings of text?