llamaindex/ParseBench
Dataset · https://huggingface.co/datasets/llamaindex/ParseBench
ParseBench Quick links: [🌐 Website] [📜 Paper] [💻 Code] ParseBench is a benchmark for evaluating document parsing systems on real-world enterprise documents, with the following characteristics: Multi-dimensional evaluation. The benchmark is stratified into five capability dimensions — tables, charts, content faithfulness, semantic formatting, and visual grounding — each with task-specific metrics designed to capture what agentic workflows depend on. Real-world enterprise… See the full description on the dataset page: https://huggingface.co/datasets/llamaindex/ParseBench.
First seen by Oriel on .
AI summary
ParseBench is a benchmark for document parsing systems, evaluating tables, charts, content faithfulness, semantic formatting, and visual grounding across 2,078 human-verified enterprise-document pages.
- The dataset covers 1,211 public documents across insurance, finance, government, and other domains, with 169,011 test rules.
- Its evaluation framework supports per-dimension scoring and pipeline comparison using GTRM, ChartDataPointMatch, rule-based text checks, formatting checks, and visual-grounding metrics.
- JSONL rules evaluate chart values, table structure, omissions, hallucinations, reading order, formatting, titles, LaTeX, code blocks, bounding boxes, and semantic classes.
- Annotations use frontier VLM auto-labeling followed by targeted human correction, while documents include PDFs, JPGs, and PNGs with varied layouts and difficulty.
- The benchmark is intended for document parsing systems supporting agentic workflows, especially where accurate extraction and auditability matter.
#document parsing #benchmarks #computer vision #ai agents
Generated by gpt-5.6-sol. Check the original for details.
Metric trends
Last 30 days · Up to 120 recorded samples per source and metric, with reposts kept separate · First and latest retained
Likes
Total likes reported by the source.
- Hugging Face Datasets: 168, latest observation ; +22 over observed period since
Downloads (30d)
Rolling 30-day downloads. A change in this value is not new downloads during the observed period.
- Hugging Face Datasets: 31,161, latest observation ; +3,156 over observed period since
Aliases
| Source | External ID | URL |
|---|---|---|
| Hugging Face Datasets | llamaindex/ParseBench | https://huggingface.co/datasets/llamaindex/ParseBench Read discussion |