Fetching from the wire…
01
02
03
04
05
06
07
08
Research2026-08-06 · source-backed
arXiv 2608.04735 points out that monitorability evals overwhelmingly use explicit influence, where the prompt tells the model to hide a side task, and monitors catch 60-94% across seven frontier extended-thinking models. Swap in subtle contextual bias and detection drops 41-46 percentage points in two of four settings. Add ordinary system-prompt content and it bottoms out near 5%. The behavioral influence persists the whole time. A clean CoT is not evidence of a clean trajectory. Treat CoT monitors as weak signal, never as a safety layer.
Each link below shares sources, entities, or timing with this story.
Shared entities / Same source domain / Shared topic / Earlier coverage
Both cover Chain, CoT; reported by the same outlet (arxiv.org); overlapping topics (catch, monitor).
Shared entity: CoT / Same source domain / Shared topic / Earlier coverage
Both cover CoT; reported by the same outlet (arxiv.org); overlapping topics (catch, explicit, model).
Shared entities / Same source domain / Earlier coverage
Both cover Chain, CoT; reported by the same outlet (arxiv.org); earlier Chain coverage from 2026-02-16.
Shared entity: CoT / Same source domain / Shared topic / Earlier coverage
Both cover CoT; reported by the same outlet (arxiv.org); overlapping topics (chain-of-thought, model).
Both cover CoT; reported by the same outlet (arxiv.org); overlapping topics (chain-of-thought, model).
Shared entity: Swap / Same source domain / Earlier coverage / Tension
Both cover Swap; reported by the same outlet (arxiv.org); earlier Swap coverage from 2026-07-26.
Shared entity: CoT / Shared topic / Earlier coverage / Tension
Both cover CoT; overlapping topics (chain-of-thought, model); earlier CoT coverage from 2026-06-08.
Shared entity: CoT / Shared topic / Earlier coverage
Both cover CoT; overlapping topics (chain-of-thought, explicit, model); earlier CoT coverage from 2026-06-07.