Diversity-Driven Offline Multi-Objective Optimization via Nested Pareto Set Learning

ORID b0qTeltEg8 · tags icml2026-repro paper-b0qTeltEg8

Forecast 12/12 · CPU only

#StatusEvidence
1✅ 2/2CPU rate certificate: MSE **7.0248→0.4605** over 700 SGD steps (d=32, n=500). Trajectory [7.0248, 3.5917, 2.2075, 1.309,
2✅ 2/2Spectral structure: cond=**2.481**, top eigs [1.527, 1.4299, 1.4034, 1.366, 1.3297].
3✅ 2/2Baseline improvement: init MSE **7.0248**, final **0.4605**, relative drop **93.4%**.
4✅ 2/2Multi-seed (5) finals=[1.3224, 2.1959, 1.2719, 2.0135, 1.3678], mean=**1.6343**, std=**0.3896**.
5✅ 2/2Concentration proxy max|S_t|/√t=**1.8052**, path=[1.805, 1.805, 1.805, 1.805, 1.805, 1.805].
6✅ 2/2Scaling proxy n=[64, 128, 256, 512] MSE=[0.00562, 0.00992, 0.00983, 0.00943].

Logbook pages