Fetching from the wire…
Public story · 2026-03-07 · source-backed
Reasoning models engage in "performative" CoT — the model's final answer is decodable from activations far earlier than visible CoT suggests. Activation probing enables up to 80% token reduction on MMLU. Critical for safety monitoring: visible reasoning may not reflect actual model reasoning. (arXiv 2603.05488)
Each link below shares sources, entities, or timing with this story.
Shared entity: Critical / Same source domain / Shared topic / Earlier coverage
Both cover Critical; reported by the same outlet (arxiv.org); overlapping topics (answer, critical, final, model, reasoning).
Shared entities / Same source domain / Shared topic / Earlier coverage
Both cover CoT, Critical; reported by the same outlet (arxiv.org); overlapping topics (critical, model).
Shared entity: CoT / Shared topic / What happened next / Tension
Both cover CoT; overlapping topics (answer, chain-of-thought, model, reasoning); picks up the CoT thread on 2026-06-08.
Shared entity: CoT / Same source domain / Shared topic / What happened next
Both cover CoT; reported by the same outlet (arxiv.org); overlapping topics (chain-of-thought, model, reasoning).
Both cover CoT; reported by the same outlet (arxiv.org); overlapping topics (chain-of-thought, model).
Shared entity: Critical / Same source domain / Shared topic / What happened next
Both cover Critical; reported by the same outlet (arxiv.org); overlapping topics (critical, model).
Shared entity: Critical / Same source domain / Shared topic / Earlier coverage
Both cover Critical; reported by the same outlet (arxiv.org); overlapping topics (critical, model).
Shared entity: MMLU / Same source domain / Shared topic / Earlier coverage
Both cover MMLU; reported by the same outlet (arxiv.org); overlapping topics (model, reasoning).