Are Candidate Models Really Needed for Active Learning?

ArXi:2605.14689v1 Announce Type: new Deep learning has profoundly impacted domains such as computer vision and natural language processing by uncovering complex patterns in vast datasets. However, the reliance on extensive labeled data poses significant challenges, including resource constraints and annotation errors, particularly in