Nemotron SFT Instruction Following Chat V3: Multi-Turn Human-AI Dialogue Data

A dataset designed to strengthen multi-turn, interactive capabilities, including open-ended chat and precise instruction following. The chat subset uses human-written prompts from sources like lmarena, lmsys, and wildchat as seed prompts, with responses generated by GLM-5 and selected via pairwise comparisons using a reward model. It was authored by NVIDIA and last updated on the platform in June 2026.

Use Cases

• Fine-tuning language models for multi-turn conversational ability based on the described chat subset.
• Training models on precise instruction following based on the dataset's stated design goal.
• Benchmarking or improving response quality in open-ended dialogue systems based on the use of pairwise comparisons for response selection.
• Studying human-AI interaction patterns using the human-written prompts from specified sources.

Strengths

• Responses are generated by the GLM-5 model and the best is selected via pairwise comparisons using a specific reward model (Qwen3-Nemotron-235B-A22B-GenRM-2603).
• Prompts are sourced from multiple human-written datasets including lmarena, lmsys, and wildchat.
• Dataset is designed with a specific focus on multi-turn, interactive capabilities.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file formats, and license information are unknown, which may limit suitability assessment.
• Data may reflect source bias inherent to the specific prompt platforms used.

Provenance

Source

NVIDIA

Collection Method

Prompts sourced from human-written datasets; responses generated by GLM-5 and selected via model-based pairwise comparison.

Freshness

Last updated 2026-06-04 04:37:43; freshness should be verified.