GPT_5.5_Distilled
- Type
- dataset
- Venue
- WithinUsAI (Hugging Face)
- Year
- 2026
- Source
- huggingface
- Access
- free
- Language
- English
- Added
- 2026-07-17T20:18:03.827512+00:00
- Verified
- 2026-07-17T20:18:03.827512+00:00
Summary
GPT_5.5_Distilled is a ~18.2k-row dataset of distilled text outputs from GPT-5.5, each annotated with a quality score (0.5-0.95) and source label. The entries are detailed technical responses covering software architecture, DevOps, CI/CD, rate limiting, Kubernetes, observability, and other engineering topics. It is intended for fine-tuning or distilling knowledge from GPT-5.5 into smaller open-source models.
Keywords
distillation gpt-5.5 sft instruction-tuning software-engineering devops quality-scored
Topics
NLP / Instruction Tuning
Research notes
- Small dataset (18.2k rows). Quality scores range from 0.5 to 0.95. Content appears to be AI-generated technical/engineering responses.