🎉 Our paper on improving robustness of SAM (Segment Anything Model) has been accepted to NeurIPS 2025!