Command Palette
Search for a command to run...
Nemotron SFT Instruction Following Chat v2 Instruction Following Dialogue Fine-Tuning Dataset
Nemotron-SFT-Instruction-Following-Chat-v2 is a synthetic dialogue dataset released by NVIDIA in 2025, designed to enhance the open-ended dialogue and instruction-following capabilities of large language models, supporting both model training and evaluation.
The dataset contains 1,998,568 samples in JSONL format, primarily consisting of text and metadata. The dialogues are generated by various large language models, covering different instruction-following and reasoning scenarios. The data is generated from synthetic dialogues, with source models including Kimi-K2-Thinking, GLM-4.6, Qwen3-235B-A22B-Thinking-2507, GPT-OSS-120b, Kimi-K2-Instruct-0905, and Qwen3-235B-A22B-Instruct-2507.
Dataset Composition
- reasoning_off: Dialogue data without reasoning mode enabled.
- reasoning_on: Dialogue data with reasoning mode enabled.
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.