chibifire/editscore-rl-train

editscore-rl-train A mirror of EditScore-RL-Data, the training data for OmniGen2 online RL with EditScore as the reward. What it is for The repo holds the prompts and source images the upstream authors used to fine-tune OmniGen2 with online reinforcement learning, with EditScore as the reward model. Each row has a task type, an edit instruction, metadata, a key and an image. rl.jsonl is the meta file. The rows are in Parquet shards under data/.… See the full description on the dataset page: https://huggingface.co/datasets/chibifire/editscore-rl-train.

<iframe src="https://silytics.si/embed/item/978100c9-a723-4938-aae8-484e8cb61997" width="420" height="190" style="border:0" loading="lazy" title="Silytics entry"></iframe>

Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 18 October 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.