Epstein Email Messages Extracted via Vision LLM from Screenshots

4,272 individual email messages extracted from screenshot images using the Qwen 2.5 VL 72B vision model. The dataset was created by 'to-be' and last updated on December 4, 2025. The source images are from the U.S. House Oversight Committee release.

Use Cases

• Benchmarking Vision LLM performance for text extraction from complex image-based documents.
• Analyzing email communication patterns based on structured data extracted from screenshots.
• Developing information retrieval systems for public records and released document collections.

Strengths

• Contains 4,272 individual email records.
• Uses a specific, advanced vision model (Qwen 2.5 VL 72B) for extraction, suggesting a focus on accuracy.
• Derived from a primary source: original JPG screenshot images from a U.S. House Committee release.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is known, but the specific data fields, file formats, and license are unknown.
• Freshness should be verified; the dataset was last updated on 2025-12 04 12:32:12.

Provenance

Source

U.S. House Oversight Committee release (screenshot images).

Collection Method

Extracted using Qwen 2.5 VL 72B vision model.

Time Range

null

Freshness

Last updated 2025-12-04 12:32:12.

Geography

null