Frontier Data Review
Cover illustration for “Reasoning Trace Quality in Chain-of-Thought Training Data”

Essay Model Capability and Data

Reasoning Trace Quality in Chain-of-Thought Training Data

Correct answers alone don't guarantee models learn the right reasoning patterns.

Priya Subramanian10 min read

Frontier Data Review

Model Capability and Data

That gate is cheap to run and feels principled, but it only checks the last token of a thousand-token document, and that's the whole problem with it.
From Reasoning Trace Quality in Chain-of-Thought Training Data