AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

How Hard Can It Be? Hardness-Aware Multi-Objective Unlearning

Jiangwei Chen, Xinyuan Niu, Rachael Hwee Ling Sim, Zhengyuan Liu, Nancy F. Chen, Bryan Kian Hsiang Low

ICML 2026 · Learning theory and algorithms › Prediction and decision losses · approach: Data attribution and recourse · 0 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use layer-specific constraints instead of a global constraint (ablation)HAMU-Q / HAMU-U · image classification unlearningforget quality and retain utilityImproved
increase Q (ablation)HAMU-Q · unlearning trajectoriesforget quality vs retain utilityMixed
increase U (ablation)HAMU-U · unlearning trajectoriesretain utility vs forget qualityMixed
increase Q (ablation)HAMU-Q · epochs to reach L_f=1.0number of epochsImproved
increase similarity mixing ratio rho to 0.75baselines (GA, KL, FT, SCRUB) · forget/retain similaritytrajectoryHurt
use HAMU instead of baselines at rho=0.75HAMU variants · hard unlearning scenariodelta L_f / -delta L_rImproved
increase rho to 1HAMU-Q / HAMU-U · unlearning trajectoriesdelta L_f / -delta L_rHurt
make forget data semantically similar to retain dataHAMU · WaterDrum-TOFUestimated hardness: 4.0e-4 -> 6.1e-4Hurt
use HAMU-Q instead of GDiffHAMU-Q · WaterDrum-TOFU LLM QAforget quality and retain utilityImproved
use smaller learning rate (ablation)HAMU-U · WaterDrum-TOFU dissimilarretain objective constraintImproved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy