Nexus: Same Pretraining Loss, Better Downstream Generalization via Common Minima

arXiv:2604.09258 · cs.LG · Submitted 2026-04-10 · Read on arXiv

cs.LG

Submitted: 2026-04-10

Updated: 2026-11-05

Code: https://github.com/ByteDance-Seed/seed-oss

Terminology

Sources

Related papers