Piccolo2 Embedding Model
model paperChinese text embedding model from SenseTime Research achieving state-of-the-art on the CMTEB benchmark across 6 tasks. Uses efficient multi-task hybrid loss training, scaled-up embedding dimensions, and MRL training for flexible vector dimensions.
Outputs 2
Piccolo2 Models
modelVariants
| Name | Parameters | Notes |
|---|---|---|
| piccolo-large-zh-v2 | — | — |
| piccolo-large-zh | — | Original v1 model |
| piccolo-base-zh | — | — |
Piccolo2: General Text Embedding with Multi-task Hybrid Loss Training
paperarXiv: 2405.06932