MTEB STS performance and using as feature extraction model

SeanLee97 / AnglE

Train and Infer Powerful Sentence Embeddings with AnglE | 🔥 SOTA on STS and MTEB Leaderboard

https://arxiv.org/abs/2309.12871

MIT License

398 stars 30 forks source link

MTEB STS performance and using as feature extraction model #4

Closed dhruv-anand-aintech closed 7 months ago

dhruv-anand-aintech commented 8 months ago

Hi, very interesting approach, and impressive results. Have you run the MTEB STS benchmark on your trained models? Would be interesting to see their performance v/s existing models with similar model size.

Thanks

SeanLee97 commented 8 months ago

Hi @dhruv-anand-aintech, thanks for following our work and thanks for your suggestion. We will test the STS performance on MTEB later.

SeanLee97 commented 7 months ago

Hi @dhruv-anand-aintech , we have uploaded our common model (UAE-Large-V1). It achieves SOTA on MTEB with an average score of 64.64. You can follow the UAE usage to use our latest models:)

dhruv-anand-aintech commented 7 months ago

congrats!