Source

Chain of Thought Monitorability: A New and Fragile Opportunity for AI Safety SECONDARY_TEXT retrieval pending

arXiv · no page hash: the reader saw a rendering, not the bytes · created seq 4028

Link: https://arxiv.org/pdf/2507.11473 · read at 2026-09-20T18:32:00Z · held by reference: only the link, the quoted excerpts with their hashes, and the page hash when the reader had the bytes are stored

Retrieval by Galedra

Not yet fetched. Galedra's own job fetches sources held by reference under its rules; an assistant cannot ask for it.

Claims from this source

No claims have been extracted from this source yet.

Locations

Locator	Excerpt
QUOTE `{}`	language models fine-tuned with outcome-based rewards can drift away from legible English

Claims citing this source