Synthetic data

Declining · emerged December 2020 · high April 2025 · +13% on the month before · 2 mentions on Trending in 14 days

Training and evaluation data generated by models or simulators instead of collected from people. Named by Wang et al., University of Washington: Self-Instruct: Aligning Language Models with Self-Generated Instructions (Dec 20, 2022). Also called Synthetic datasets, Model-generated data, Self-instruct.

Index 0 to 100, monthly

Signals

Signal	Value	Reads as
Hacker News stories, last 12 months42 stories, -39% on the 12 months before	42 stories, -39% on the 12 months before	Decline
arXiv papers, last 12 months586 papers, -10% on the 12 months before	586 papers, -10% on the 12 months before	Decline
Wikipedia views, August 20261,906, -43% on a year before	1,906, -43% on a year before	Decline
Against its high-54% from the high of April 2025	-54% from the high of April 2025	Decline
Product releases naming it2 since Sep 1, 2025	2 since Sep 1, 2025	Adoption
GitHub repositories15,793	15,793	Adoption

Companies

Name	Mentions
Oracle	1
Microsoft	1

Releases

2 releases since Sep 1, 2025

Name	Date
Select AI for JavaOracle · Sep 22, 2026 3 days ago	Sep 22, 20263 days ago
Cloud evaluations with the Microsoft Foundry SDKMicrosoft · Aug 1, 2026 7 weeks ago	Aug 1, 20267 weeks ago

Papers

Name	Date
Nemotron-4 340B Technical ReportNVIDIA · Jun 17, 2024 2 years ago	Jun 17, 20242 years ago
Phi-3 Technical Report: A Highly Capable Language Model Locally on Your PhoneMicrosoft · Apr 22, 2024 2 years ago	Apr 22, 20242 years ago

GitHub projects

15,793 repositories

Name	Created	Stars
Kiln-AI/KilnBuild, Evaluate, and Optimize AI Systems. Includes evals, RAG, agents, fine-tuning, synthetic data generation, dataset management, MCP, and · Jul 23, 2024 2 years ago	Jul 23, 20242 years ago	5,079
Belval/TextRecognitionDataGeneratorA synthetic data generator for text recognition · Jul 1, 2017 9 years ago	Jul 1, 20179 years ago	3,695
sdv-dev/SDVSynthetic data generation for tabular data · May 11, 2018 8 years ago	May 11, 20188 years ago	3,565
argilla-io/distilabelDistilabel is a framework for synthetic data and AI feedback for engineers who need fast, reliable and scalable pipelines based on verified · Oct 16, 2023 2 years ago	Oct 16, 20232 years ago	3,397
hitsz-ids/synthetic-data-generatorSDG is a specialized framework designed to generate high-quality structured tabular data. · Aug 10, 2023 3 years ago	Aug 10, 20233 years ago	2,439
NVIDIA-NeMo/DataDesigner🎨 NeMo Data Designer: Generate high-quality synthetic data from scratch or from seed data. · Oct 16, 2025 11 months ago	Oct 16, 202511 months ago	2,278

Movement	Stage
Small language modelsCompact models, a few billion parameters or less, trained to be good enough and cheap to run anywhere.	Peak
Reasoning modelsModels trained with reinforcement learning to think step by step, spending more compute at answer time.	Plateau
LLMOps and evalsMeasuring, tracing and monitoring model-backed software: eval suites, LLM observability and prompt management.	Peak

History

Date	What changed
Sep 24, 2026	Tracking started: emerged December 2020, declining, 3 other names recorded

Sources: Curve: Hacker News story titles, Wikipedia pageviews and arXiv papers by month, refreshed Sep 24, 2026. Related entities from the fru.dev sites' public APIs, matched by name. How stages work.