Skip to content

Dataset · SargeDev

System One fast judge model training corpus

SargeDev/jev-distill-corpus-v3

An English-language dataset of 740,957 examples for distilling compact classifiers that evaluate states and decisions in a single pass.

What's inside

Contains 740,957 English rows with calibrated typed decisions (noul, choice, and score schemas). A significant portion (498,010 rows) consists of synthetic operational scenarios across 53 domains, including business, engineering, and agent workflows. The data is designed for training small local models to return probability distributions without slow reasoning.

What you can do with it

  • Distilling fast judge models based on BERT or compact Qwen
  • Training classifiers for instant evaluation of AI agent actions
  • Calibrating probabilities in business decision-making processes
Tasks
Text classification
Languages
English
Size
100K–1M rows
Downloads this month
1.9K
Likes
50

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("SargeDev/jev-distill-corpus-v3")