• GBU_28@lemm.ee
    link
    fedilink
    English
    arrow-up
    3
    ·
    6 months ago

    It’s use in hybrid search CRAG and discussion systems as a human in the loop augmentation is quite valuable. It saves analyst time and streamlines further review. In a sufficiently adversarial system where multiple models are vetting the responses for sanity with back checks, it’s quite performant, having decent recall and accuracy.