Patent Reexamination and Appeal Legal Evidence Retrieval Benchmark
#legal evidence retrieval #disputed issue identification #evidence-document matching #patent reexamination #patent appeal proceedings #intellectual property dispute analysis
- 5000
- 0.2GB
- JSON
- CATL
MOBIUSI INC
Updated:2026-10-19
AI Analysis & Value Prop
This benchmark contains legal documents from patent reexamination or appeal proceedings, including appeal briefs, responses, and decisions. It defines evidence retrieval tasks around disputed issues, decision reasoning, and adopted arguments, with annotations identifying direct support in the decision text and links between arguments and evidence. It supports evaluation of evidence localization, retrieval ranking, citation accuracy, and knowledge grounding in dense patent-law reasoning, as well as the development and validation of legal retrieval-augmented generation systems.
Dataset Insights
Sample Examples
{}
Technical Specifications
| Field | Type | Description |
|---|---|---|
| file_name | string | File name |
| evidence_spans | array | Gold-standard supporting evidence organized as retrievable, locatable, and comparable text spans. |
| evaluation_task | object | Input retrieval tasks focused on disputed issues, decision reasoning, or adopted arguments. |
| normalized_task | object | Standardized disputed issues and retrieval targets derived from the original task and document content. |
| gold_annotations | object | Human annotations of decision outcomes, direct supporting evidence, and adopted arguments for evaluating retrieval and evidence grounding. |
| source_documents | array | Original documents and text from patent reexamination or appeal proceedings, including appeal briefs, responses, and decisions. |
| retrieval_evaluation | object | Metrics calculated from gold-standard evidence and adopted arguments to assess system retrieval, citation, and evidence-linking performance. |
| retrieval_submission | object | Ranked results, answers, and citations submitted by a retrieval or generation system for evaluation. |
| argument_evidence_links | array | Links between adopted arguments and their direct supporting evidence for evaluating knowledge-grounding capability. |
Compliance Statement
| Authorization Type | Proprietary - Commercial AI Training License (No Redistribution) |
| Commercial Use | Requires exclusive subscription or authorization contract (monthly or per-invocation charging) |
| Privacy and Anonymization | No PII, no real company names, simulated scenarios follow industry standards |
| Compliance System | Compliant with China's Data Security Law / EU GDPR / supports enterprise data access logs |
Frequently Asked Questions
- What documents and tasks does the benchmark include?
- The corpus covers appeal briefs, responses, and decisions from patent reexamination or appeal proceedings. Tasks retrieve evidence related to disputed issues, decision reasoning, and adopted arguments, with normalized retrieval targets and evidence spans.
- What capabilities can the gold annotations evaluate?
- Human annotations include decision outcomes, direct support in the decision text, and adopted arguments, with links between arguments and evidence. They can be used to assess retrieval ranking, evidence localization, answer citations, and grounding quality.
- Which systems and research use cases is it suitable for?
- It is suitable for patent-law retrieval and for developing or evaluating legal retrieval-augmented generation systems. It can also support comparisons of issue identification, direct-evidence recall, and citation accuracy in legal reasoning.
- How can the data be obtained, and what usage restrictions apply?
- Refer to the dataset product page for availability and license terms. Before use, verify the permitted scope for document content, privacy requirements, and redistribution restrictions. Evaluation results should also be reviewed in the context of the intended legal task.
Can't find the data you need?
Post a request and let data providers reach out to you.
Get this Dataset
Verified for Enterprise UseCite this Work
@dataset{Mobiusi2026,
title={Patent Reexamination and Appeal Legal Evidence Retrieval Benchmark},
author={MOBIUSI INC},
year={2026},
url={https://www.mobiusi.com/datasets/549a7f09660085a787a43f157780296d?dataset_task_cate_id=10},
urldate={2026-10-15},
keywords={patent reexamination dataset,patent appeal documents,legal evidence retrieval,patent decision evidence,legal RAG benchmark,legal issue retrieval,argument evidence linking,patent law documents},
version={1.0}
}Using this in research? Please cite us.
placeholder
placeholder
placeholder
placeholder
placeholder
placeholder
placeholder