Early-Signal Detector

method · first seen 4 Oct 2026

multi-objective alignment

Also: MOA

Training language models to satisfy multiple, possibly conflicting, human preference objectives and trace the Pareto front.

Mentions per week

0119 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

reinforcement learning from human feedbackminimax preference optimization