Ex-OpenAI Researcher: Poor Pretraining Stacks Produce Misleading 'Useless Layers' Claims

Former OpenAI researcher @aidanclark argued that widespread claims about the uselessness of Transformer's later layers don't hold on well-built pretraining stacks, showing how poor training pipelines can produce misleading research conclusions.

2026-10-16 ~ 2026-10-16 · 2 related posts

• Researcher: 'later transformer layers are useless' is an artifact of weak pretraining stacks — _aidan_clark_ · 2026-10-16
• Researcher: 'later transformer layers are useless' is untrue on good pretraining stacks — _aidan_clark_ · 2026-10-16