Genμ 2.0 (2026) — Leaderboard

Leaderboard

Competition standings — 2026 edition

Verified submissions are ranked on this website. Face and Speech rankings are determined only by ERB; supporting preservation and quality measurements are shown separately. The Concept Track continues to use ERR.

StatusChallenge open — submissions are accepted until 10 July 2026.

TracksTask 1 (Face Unlearning), Task 2 (Speech Unlearning), Task 3 (Concept Unlearning).

Leaderboard updatesVerified results are posted here as evaluation progresses.

Current status

The challenge is open. Submit through the track-specific Google Forms by 10 July 2026. Submitted weights and repositories are verified, qualifying entries are added below, and final results are declared on 25 July 2026.

Face Track Form — link pending Speech Track Form — link pending Concept Track Form — link pending

Ranking rule

Face and Speech: rank is determined only by \(\mathrm{ERB}=2(EA)(RA)/(EA+RA)\), where \(EA=100-FA\). Other measurements are displayed for transparency but do not affect rank. Concept: the original ERR score is used. Full definitions are on the Evaluation page.

Leaderboard — Task 1: Face Unlearning Track

Rank	Team	Model / checkpoint	ERB ↑	FA ↓	RA ↑	GP ↑	AR ↑	Updated
No verified entries yet. Rankings will be added after evaluation is completed.

Leaderboard — Task 2: Speech Unlearning Track

Rank	Team	Model / checkpoint	ERB ↑	FA ↓	RA ↑	PPS ↑	SAR ↑	Updated
No verified entries yet. Rankings will be added after evaluation is completed.

Leaderboard — Task 3: Concept Unlearning

Rank	Team	Model / checkpoint	ERR Score ↑	Forgetting \(\bar{A}_{\mathrm{fgt}}\) ↑	Retention \(\bar{A}_{\mathrm{ret}}\) ↑	Robustness \(\bar{A}_{\mathrm{adv}}\) ↑	Weight \(\Delta\) ↓	Updated
No verified entries yet. Rankings will be added after evaluation is completed.