Contrastive continue-training of v1 with mined hard negatives. Use query: / passage: prefixes + mean pooling + L2 normalize.
query:
passage:
Files info
Base model
