Phylo Tests Suggest Claude Trained Heavily on Protein Sequences
Independent tests by Xinming Tu of the Phylo team indicate that the latest Claude model was likely trained on large amounts of protein sequence data, correctly predicting obscure protein sequences across 668 bacterial and archaeal proteins, while other frontier models show no such signs.
2026-11-18 ~ 2026-11-18 · 2 related posts
- Third-party test suggests Claude was trained on protein sequences at scale — AllThingsApx · 2026-11-18
- Claude was trained on protein sequences at scale, Phylo test finds — no other frontier model shows it — AllThingsApx · 2026-11-18