facts: branch points localize where the feature choice is causally responsible, as the paper puts it, rather than proving that features drive output quality
facts: layer 10 is notation and entity tracking (formatting is layer 20), most-active steering slightly hurts BBQ Ambig, seven benchmarks, the critic curves rise, GRPO is critic-free rather than preference-based, the critic bottleneck is inferred, and HarmBench and XSTest get the citations they never had
facts: L10/12607 back in the Layer 10 column with layer 20 no longer credited for its logical structure, two citations that rendered as raw keys, and most-active steering scoped to BBQ
