UNMASK Auto-Detects and Fixes Spurious Shortcuts in Text Models
UNC-ChapelHill · hf · 2026-08-17
UNC-Chapel Hill introduces UNMASK, a framework that automatically discovers and mitigates spurious correlations in text classifiers. It uses causal verification and group-based reweighting without manual annotations.
More from Research
- IBM Open Sources Docling-Graph: Converting PDFs to Knowledge Graphs — aigleeson · 2026-08-17
- ZipSplat: Fewer Gaussians, Better 3D Scene Reconstruction — rsasaki0109 · 2026-08-17
- Petition calls for mandatory quantization labels in model评测 posts — Su1tz · 2026-08-17
- A Pathway to General-Purpose Scientific AI: New Multimodal Benchmark — SciKnowOrg · 2026-08-17
- Subjective Logic: A Foundational Book on Reasoning Under Uncertainty — FrnkNlsn · 2026-08-17
- Study: Simple methods achieve high accuracy in known protein interaction prediction — anshulkundaje · 2026-08-17