NVIDIA 发布开源 midtraining 配方，在 SFT 不变的前提下加入一个中间训练阶段，Nemotron 3.5 Lightning 的 SWE-bench Multilingual 从 55.8 升至 62.8，SWE-bench Verified 从 60.7 升至 64.5，SWE-bench Pro 从 32.0 升至 35.5。
Great paper from NVIDIA sharing an open midtraining recipe for agentic coding.
NVIDIA kept SFT identical and added one midtraining stage. SWE-bench Multilingual went from 55.8 to 62.8 on Nemotron 3.5 Lightning.
The stage sits between pretraining and post-training. It trains at 256K context on 500B tokens of natural software-engineering data, such as packed repositories, pull requests, commit histories and build logs, plus a small share of agent traces.
On the same model, SWE-bench Verified rose from 60.7 to 64.5 and SWE-bench Pro from 32.0 to 35.5. On the larger Nemotron 3.5 Super, DeepSWE rose from 11.7 to 15.7.
They also explain how they evaluated the recipe. Base-model proxy metrics did not predict post-training results, so every data choice was scored with short fixed SFT runs.
If you train coding agents, the recipe and its ablations are published.
https://arxiv.org/abs/2610.22019
Chat with Paper: https://academy.dair.ai/papers/priming-base-models-for-agentic-post-training-an-open-midtraining-recipe-2610.22019
来源：DAIR.AI · x.com