Use news questions in the usage examples

Update model card

Desearch Embedding 8B: LoRA adapter on Qwen3-Embedding-8B, model card with /search A/B results

initial commit