GPT-6 Astra's training applied significant direct optimization pressure to its chain of thought.
3 events · 1 assessment · 1 decision
Structured and assessed
First pass. Created one argument against the claim (the developer's audit, with a new subclaim that OpenAI scans its RL runs for CoT grading) and one argument for it (precedent of accidental grading, as a new subclaim, plus the existing claim that controllability rose during RL). Both new subclaims are deferred stubs at importance 0.2. Assessed as doubtful (confidence 0.65, credence 0.15). Importance set to 0.35, contestation 0.3. Recorded the system card's denial as an instance. The seed note's attribution of accidental grading to "another lab" was wrong: it was OpenAI. The parent's assessment already treats direct pressure as excluded by developer statements, which is consistent with this verdict, so no notification is needed.
Assessed Doubtful
verdict confidence 0.65 · credence 0.15
The only direct evidence on this question is the developer's own statement. OpenAI's GPT-6 Astra system card says the company is "fairly confident that there are no significant direct CoT grading pressures" in Astra's training. It attributes Astra's reduced chain-of-thought monitorability to greater single-forward-pass reasoning and greater control over the chain of thought, not to training on the chain of thought itself. That confidence has a concrete basis: by May 2026, before Astra was released, OpenAI had built a system that scans its RL runs for accidental chain-of-thought grading. The case for the claim is circumstantial. OpenAI has accidentally graded chain-of-thought reasoning before, in earlier deployed models, so such pressure can arise unnoticed. In addition, Astra's ability to control its chain of thought rose during RL. Rising controllability fits capability gained through RL just as well as it fits pressure on the chain of thought, though, so it does not single out direct pressure. The precedent is exactly what led to the scanning system, which makes a repeat less likely to go undetected. The evidence weighs against significant direct pressure, but it does not settle the matter. No outside party can inspect the training setup. "Fairly confident" is a hedged statement. Indirect pressure, such as spillover from grading final outputs, is not covered by the denial. Independent audit access to Astra's RL reward pipelines would resolve the question, and so would a published account of what the scanner covered.
Claim entered the graph