Full talk · edited transcript

Embedders 2025: A New Vector in Text Representation

Roman Solomatin24 May 2025

Roman Solomatin, X5 Tech (AiRUN team, chatbots and search; also a researcher at DeepPavlov, a master's student at AITH, an MTEB maintainer and one of the authors of ruMTEB, the Russian-language part of the benchmark)

The talk is an overview: what encoders and sentence models are, which up-to-date embedding models have appeared recently, and how their quality is measured. The speaker promised not to talk about LLMs but ended up doing so, because LLMs have become the basis of new embedders.

Continue reading — it's free

Google or a one-time email code. No password.