ICCV 2025: Two Workshop Papers

🎉🎉🎉 AIML@K contributes two workshop papers to ICCV 2025!

Kudos to Jaeheun, Jaehyuk, Yeajin, Bosung, and Suhyun!

OPC: One-Point-Contraction Unlearning Toward Deep Feature Forgetting presents a new machine unlearning algorithm that enforces deep feature forgetting by contracting representations of data to be forgotten, achieving stronger resistance to recovery and gradient-inversion attacks than existing unlearning methods.

IPPRO: Importance-based Pruning with PRojective Offset for Magnitude-indifferent Structural Pruning introduces a novel magnitude-indifferent structured pruning method, named IPPRO, which embeds filters into a projective space and defines a new importance score (PROscore) based on gradient movement, enabling loss-minimal pruning beyond the “size matters” convention.