Mixed edge case datasets for cross-workflow testing.

The testing job

Test the seams between document types.

Real workflows rarely receive one perfectly uniform input. Mixed cases help you inspect what happens when formats, labels, context, and expected actions change within the same testing run.

• Different document formats in one release
• Conflicting or incomplete context across files
• Unexpected fields, ordering, and naming
• Cases that cross parsing, retrieval, and classification

Useful checks

Follow the case across the whole workflow.

Use mixed edge cases for ingestion checks, routing, retrieval, extraction, evaluation, and regression testing. The goal is to reveal assumptions that stay hidden when every example follows one schema.

• Does the pipeline identify the right document type?
• Does it preserve context while formats change?
• Are unsupported inputs surfaced safely?
• Can a reviewer trace the final result to each source?

Deniable approach

Human judgement defines the difficult combination.

Human input and creative scenario design determine which edge conditions belong together. The collection is made for internal development, QA, evaluation, training, and fine-tuning, with the testing purpose documented rather than implied.

Catalogue note

Choose a pack that matches your workflow.

Each catalogue release will list its included formats, metadata, and document count. Fixed packs use ZIP delivery, while authenticated API retrieval uses prepaid credits for selected documents.