metadata
license: apache-2.0
language:
- zh
tags:
- text2vec
- feature-extraction
- sentence-similarity
- transformers
pipeline_tag: sentence-similarity
based on https://github.com/shibing624/text2vec
- using all 5 tasks' datasets
- base model is hfl/chinese-lert-base
- max_seq_length = 256
spearman result:
STS-B | ATEC | BQ | LCQMC | PAWSX |
---|---|---|---|---|
67.60 | 42.67 | 61.61 | 77.79 | 57.54 |