Interventional Training Reduces Causal Bypass in In-Context Introspection.

arXiv:2610.38911v1 Announce Type: cross Abstract: Introspection, the ability of a system to access and reason about its internal signals, is crucial for language model (LM) self-consistency and scalable interpretability. Prior work argues that LMs possess intr