Sources
Loading...
Additional media
Loading...
DeepNewz, mobile.
People-sourced. AI-powered. Unbiased News.
Answer.ai has released a new model, ColBERT-Small-V1, developed by bclavie. This small-sized model, with just 33 million parameters, outperforms the larger bge-base model on the BEIR benchmark. The ColBERT-Small-V1 is designed for late interaction retrieval tasks and is praised for its efficiency in indexation, classification, generation, and parsing. The model is compatible with Vespa Engine, offering improved performance over previous models.
Here is how you use the small but mighty answerai-colbert-small-v1 model from @answerdotai with @vespaengine! Thank you, @bclavie, for your support and for pushing a new ColBERT model checkpoint that outperforms the older (and larger) one. https://t.co/MA2KTlzOc1
Small but Mighty: Introducing answerai-colbert-small https://t.co/68jkuBxsB9
Super cool work. Amazing to see how far the late interaction retrieval architecture of ColBERT still gets us today, with tiny models! https://t.co/IOQh8vIKKh
