Abstention Direction in Large Language Models: Mechanistic Interpretability of Abstention Across Query Ambiguity, Knowledge Limits, and Safety Alignment

arXiv:2611.20441 · cs.CL · Submitted 2026-11-16 · Read on arXiv

cs.CL

Submitted: 2026-11-16

Updated: 2026-11-16

Terminology

Related papers