Paper recorded by Signals 4 on 2026-09-24 in cs.AI. Abstract reproduced from arXiv; link to the original below.
Published 2026-09-24 on arXiv · recorded by Signals 4 on 2026-09-25
Category: cs.AI · 人工智能 · first seen 2026-09-25
Pre-logit steering adapts a frozen language model to a test-time reward by adding vectors to its final hidden states. Unregularized reward optimization can substantially alter the output distribution and degrade generation quality. We propose Minimally Invasive Steering Vector Optimization (MISVO), which penalizes interventions using the local KL geometry of the induced token distribution. The res