TE
TechEcho
Home24h TopNewestBestAskShowJobs
GitHubTwitter
Home

TechEcho

A tech news platform built with Next.js, providing global tech news and discussions.

GitHubTwitter

Home

HomeNewestBestAskShowJobs

Resources

HackerNews APIOriginal HackerNewsNext.js

© 2025 TechEcho. All rights reserved.

Yet another open-source LLaMA-Based sentence embeddings

4 pointsby yokeeover 1 year ago

1 comment

yokeeover 1 year ago
Recently, sentence embeddings play an important role in information retrieval, especially in LLM-augmented retrieval. AnglE (<a href="https:&#x2F;&#x2F;github.com&#x2F;SeanLee97&#x2F;AnglE">https:&#x2F;&#x2F;github.com&#x2F;SeanLee97&#x2F;AnglE</a>) is an open-source sentence embedding model. It shows good performance in Semantic Textual Similarity (STS) tasks and currently supports generating LLaMA-7B embeddings, which are state of the art in STS, as well as BERT-based embeddings. It also supports two languages: English and Chinese. LLaMA&#x27;s powerful multilingual capability makes it easy to fine-tune the model for other languages.