ParseBench-2: Severity-Aware Evaluation of Document Parsing
Paper · https://arxiv.org/abs/2611.17400
arXiv:2611.17400v1 Announce Type: new Abstract: Document-parsing metrics can give high scores to outputs that change important facts and penalize changes that preserve meaning. People can often judge these consequences more easily than they can design a metric to measure them. We introduce a method for developing parsing metrics from severity policies that specify how each error type should affect the score. We express these policies through test scenarios that specify how a metric should behave. We then optimize the metric implementations against these fixed expectations, using coding agents to refine comparison rules, error detectors and weights under human review. This makes stated human priorities an explicit objective of metric design. We apply this method in ParseBench-2, a comprehensive document-parsing benchmark that combines broad capability coverage with metrics calibrated to error severity. It covers 4,317 pages across six dimensions of parsing quality: text, formatting, tables, charts, visual grounding, and forms. We use detailed annotations and targeted difficult cases to check for changed facts, broken structural relationships, and missing content. We take the lower of a parsing score that measures content recovery and a severity cap that penalizes detected errors, so recovering most of a page cannot hide a serious error. Across six severity test suites, our metrics more closely match the expert judgments expressed by the scenario targets than existing metrics and penalize at most 3.4% of the tested benign variations. We conduct a comprehensive evaluation spanning frontier vision-language models, specialized document parsers and open-weight systems, comparing their overall rankings, performance across parsing capabilities and tradeoffs between accuracy and cost.
First seen by Oriel on .
AI summary
ParseBench-2 introduces severity-aware document-parsing metrics that align scoring with human judgments and prevent serious errors from being hidden by high content-recovery scores.
- The method encodes severity policies as test scenarios, then uses coding agents under human review to refine comparison rules, error detectors, and weights.
- The benchmark covers 4,317 pages across text, formatting, tables, charts, visual grounding, and forms.
- Scores take the lower of content recovery and a severity cap, ensuring detected serious errors limit overall results.
- Across six severity test suites, the metrics matched scenario-based expert judgments more closely than existing metrics while penalizing at most 3.4% of benign variations.
- The evaluation compares frontier vision-language models, specialized document parsers, and open-weight systems across capabilities, rankings, accuracy, and cost.
#document parsing #benchmarking #vision-language models #evaluation metrics
Generated by gpt-5.6-sol. Check the original for details.
Aliases
| Source | External ID | URL |
|---|---|---|
| arXiv AI Papers | oai:arXiv.org:2611.17400v1 | https://arxiv.org/abs/2611.17400 Read discussion |