| language: | |
| - en | |
| tags: | |
| - embeddings | |
| - information-retrieval | |
| - ramana-maharshi | |
| - contrastive-learning | |
| base_model: SriRamanaAtmic/AtmicEmbeddingv2 | |
| license: mit | |
| pipeline_tag: sentence-similarity | |
| # AtmicEmbedding (English query→passage) | |
| Contrastive continue-training of AtmicEmbeddingv2 on the expert_pass Q&A dataset, with up to 3 mined hard negatives per anchor. Use `query: ` / `passage: ` prefixes + mean pooling + L2 normalize. | |