Teknium: Hermes Shows Both Memory and Generalization, ~15% Over Fine-tuning

Teknium commented on Unify's research on the Hermes model's memory mechanism, noting that the results show Hermes both memorizes and generalizes, achieving roughly 15% better generalization than a fine-tuning baseline of about 28%.

2026-11-12 ~ 2026-11-12 · 2 related posts

• Teknium on Unify's memory research: Hermes shows both memorization and ~15% gain over finetuning — Teknium · 2026-11-12
• Teknium: Hermes memory research shows ~15% better generalization than finetuning — Teknium · 2026-11-12