Command Palette
Search for a command to run...
Nemotron RL Agentic SWE Pivot v1 Software Engineering Agent Reinforcement Learning Dataset
Nemotron RL Agentic SWE Pivot v1 is a dataset released by NVIDIA in 2026 for training and validating real-world software engineering agents, with the related research paper available at Training Software Engineering Agents and Verifiers with SWE-Gym. It is designed to support model fine-tuning using reinforcement learning from verifiable rewards (RLVR) within the NeMo Gym environment.
This dataset is a restructured version of the SWE-Gym and R2E-Gym datasets, specifically tailored to fit the NeMo Gym input format, and has been applied in the development of NVIDIA's Nemotron series models. It is automatically converted from SWE-Gym and R2E-Gym, primarily built on GitHub Issues, and aims to adapt to the NVIDIA NeMo Gym framework for post-training of large language models as well as reinforcement learning training and validation of software engineering agents.
Dataset Composition
- responses_create_params: Parameters related to response generation.
- agent_ref: Reference information related to agents.
Build AI with AI
From idea to launch — accelerate your AI development with free AI co-coding, out-of-the-box environment and best price of GPUs.