As of Oct 9, Datology launched Curation Studio, a guided, configurable product that cleans, calibrates, augments with synthetic data and composes training data inside the customer's VPC.
DatologyAI published this on Oct 9, 2026 (source: datologyai.com).
Datology launched Curation Studio, which packages its data-curation research into a guided, configurable product that takes raw data through four stages: clean, calibrate, create synthetic data, and compose training sets. It runs inside the customer's VPC, and the first release covers LLM data across web, math, code and multilingual. According to the company, a 30B-A3B model trained on its curated data scored 46.8% versus 37.7% for a baseline built from 39 public datasets (both pretrained on 1 trillion tokens), and models matched the largest baseline with 6x less training compute.
Items older than 72 hours never go in Top stories and show their original date.