xiongyq/SCOPED-Train
SCOPED training data Paper: Skill-Decoupled On-Policy Expert Distillation for Unified Vision-Language Models Code: GitHub Four separate capability configurations, each with the original train/test split. Images are embedded in the Parquet files; no original machine paths or private skills are required. Configuration Task Train Test capability1 Fine-grained perception, counting and grounding 9499 501 capability2 OCR, charts and table structure 9500 500 capability3… See the full description on the dataset page: https://huggingface.co/datasets/xiongyq/SCOPED-Train.
| Author / organization | xiongyq |
|---|---|
| Downloads | 165 (as reported by Hugging Face) |
| Likes | 1 |
| Created | 6 October 2026 |
| Last updated at source | 17 October 2026 |
Popularity, last 4 days
165 now (+11 since 2026-11-04). Counts as reported by the source, snapshotted daily by Silytics.
Connected in the knowledge graph
- mradermacher/SCOPED-4B-GGUFModel · used to train
- Skill-Decoupled On-Policy Expert Distillation for Unified Vision-Language ModelsPaper · cites
- mradermacher/SCOPED-8B-GGUFModel · used to train
- xiongyq/SCOPED-4BModel · used to train
Related papers
More from xiongyq
- xiongyq/SCOPED-EvalDataset
Silytics shows public metadata from Hugging Face and links to the original. Information may be incomplete or out of date. Last refreshed 9 November 2026. Licenses and terms of the original apply to the work itself. Not financial, legal or professional advice.