LightOn Releases EDiTh, a Synthetic Enterprise Dataset for RAG Benchmarking
IgorCarron · x · 2026-08-22
LightOn released EDiTh, a synthetic enterprise corpus for benchmarking retrieval in RAG systems. It simulates a fictional multinational group, Véracier Industries, with 1,004 PDFs across 6 languages and 36 ground-truth use cases.
Key Features:
- Fully synthetic, avoiding privacy issues of real data.
- Designed to demonstrate what modern retrieval pipelines deliver.
- Available on Hugging Face.
More from Apps
- Zoom pivots from meeting app to creator economy infrastructure — charitychaste · 2026-08-22
- Krea2 Launches Insanely Simple LoRA Training — PutridExplanation394 · 2026-08-22
- Open Source LogoCreator Generates Brand-Ready Logos and Full Kits Using Flux — tom_doerr · 2026-08-22
- Why AI background removers fail on wreaths and the chroma-key alternative — ShimmerThiyaelx · 2026-08-22
- Top 3 Rankings Equal Citations in Google AI Overviews — gaganghotra_ · 2026-08-22
- Minimax H3 Prompt Composer Update Fixes Camera Planning Bug — Brad12d3 · 2026-08-22