NVIDIA Open-Sources Midtraining Recipe to Boost Agentic Performance

NVIDIA published an open midtraining recipe that inserts a stage between pretraining and posttraining, using 500B tokens at 256K context with repository, PR, and commit-history data, boosting SWE-bench Multilingual by 7 points.

2026-10-17 ~ 2026-10-17 · 2 related posts

• NVIDIA open-sources midtraining recipe: 500B tokens lift Nemotron agentic benchmark scores — denisyarats · 2026-10-17
• NVIDIA's open midtraining recipe lifts SWE-bench Multilingual from 55.8 to 62.8 with 500B tokens — dair_ai · 2026-10-17