Apple Research: Low-Impact Points Reduce Unlearning Costs
Apple ML Research · rss · 2026-07-17
This Apple ML Research paper addresses the challenge of "unlearning." When removing specific data points from a trained model, existing methods typically treat all samples in the forget set equally, which isn't always necessary.
The authors suggest using influence functions to analyze how training samples affect model outputs. By identifying low-impact data points, they can narrow down the scope of samples requiring processing, thereby reducing the computational costs of unlearning. The paper includes a comparative analysis across both language and vision tasks.
More from Research
- A parody prompt asks for a planetary crystal factory inventory and parity report — Promptmethus · 2026-07-21
- PRA hits new image-generation SOTA with 511M parameters and FID 1.94 — jiqizhixin · 2026-07-21
- SUFLECA shows NOC-based correspondence can improve CAD-to-image alignment — ducha_aiki · 2026-07-21
- OpenAI-style autonomous researchers could become real scientific collaborators — Promptmethus · 2026-07-21
- Soft Clamp cuts tool-call overuse in multi-teacher distillation, from 13.7% to 9.0% — antgroup · 2026-07-21
- ShotPlan adds learnable planning tokens for cinematic multi-shot video generation — Tele-AI · 2026-07-21