ICPSL Framework for In-Context Preference-Conditioned Pareto Set Learning
October 16, 2026
ICPSL introduces a foundation-style hypernetwork that maps specified preferences directly to Pareto optimal solutions using limited offline data. This method amortizes both preference-wise search and problem-specific model construction, bypassing the need for repeated training or search.
HOW THIS AFFECTS YOU
●
researcherYou can use this framework to handle multi-objective optimization tasks where objective evaluations are unavailable during deployment.