This repository contains the sft training corpus produced by the original
RLAD abstraction-generator pipeline. The examples are stored in train.jsonl
with user/assistant messages; metadata.json records generation or rejection
statistics.
Base model: Qwen/Qwen3-1.7B
RLAD source commit: 0325ea848734b26141f3896ca28ff58b5dd94736
Source data: agentica-org/DeepScaleR-Preview-Dataset
Review the source dataset and base-model licenses before… See the full description on the dataset page:
https://huggingface.co/datasets/zjhhhh/rlad-original-absgen-sft-data.
