Early-Signal Detector

method · first seen 8 Oct 2026

preference alignment

Tuning a model with pairs of preferred and dispreferred outputs to reduce undesired behaviors.

Mentions per week

0219 May11 Aug10 NovPapersCodeDiscussion

PapersCodeDiscussion

Earliest sources

Latest sources

Mentioned together with

ultrasound report generationdifferential privacyDP-SGDreinforcement learning with verifiable rewards