AnglE📐: Angle-optimized Text Embeddings
It is Angle 📐, not Angel 👼.
🔥 A New SOTA Model for Semantic Textual Similarity!
Github: https://github.com/SeanLee97/AnglE
STS Results
| Model | ATEC | BQ | LCQMC | PAWSX | STS-B | SOHU-dd | SOHU-dc | Avg. |
|---|---|---|---|---|---|---|---|---|
| ^shibing624/text2vec-bge-large-chinese | 38.41 | 61.34 | 71.72 | 35.15 | 76.44 | 71.81 | 63.15 | 59.72 |
| ^shibing624/text2vec-base-chinese-paraphrase | 44.89 | 63.58 | 74.24 | 40.90 | 78.93 | 76.70 | 63.30 | 63.08 |
| SeanLee97/angle-roberta-wwm-base-zhnli-v1 | 49.49 | 72.47 | 78.33 | 59.13 | 77.14 | 72.36 | 60.53 | 67.06 |
| SeanLee97/angle-llama-7b-zhnli-v1 | 50.44 | 71.95 | 78.90 | 56.57 | 81.11 | 68.11 | 52.02 | 65.59 |
^ denotes baselines, their results are retrieved from: https://github.com/shibing624/text2vec
Usage
from angle_emb import AnglE
angle = AnglE.from_pretrained('SeanLee97/angle-roberta-wwm-base-zhnli-v1', pooling_strategy='cls').cuda()
vec = angle.encode('你好世界', to_numpy=True)
print(vec)
vecs = angle.encode(['你好世界1', '你好世界2'], to_numpy=True)
print(vecs)
Citation
You are welcome to use our code and pre-trained models. If you use our code and pre-trained models, please support us by citing our work as follows:
@article{li2023angle,
title={AnglE-Optimized Text Embeddings},
author={Li, Xianming and Li, Jing},
journal={arXiv preprint arXiv:2309.12871},
year={2023}
}