Nemotron-SFT-SWE-v3 is a software engineering instruction fine-tuning dataset released by NVIDIA in 2026. Related research papers include... Training Software Engineering Agents and Verifiers with SWE-GymThe aim is to improve the code understanding and repair capabilities of large language models in SWE-Bench style tasks. It is widely used in software engineering agent training, code repair and generation tasks, SWE-Bench style evaluation, and research and optimization of large models in code understanding and multi-step tool invocation capabilities. This dataset contains 237,970 samples. The data comes from agent trajectories collected using various agent frameworks such as OpenHands, SWE-agent, and mini-SWE-agent, and is collected and labeled through automated and synthetic methods.