Documentary Production
EP03: Ghosts in the Machine — When Machines Hallucinate
TLDR Episode 3 of PAPER TRAIL confronts the pipeline's most dangerous failure mode: phantom entities generated by OCR hallucination. Of the top 25...
#VLM
10 investigations
Every Epstein Revealed investigation tagged #VLM — reporting built on court filings, financial records, flight logs, and government document releases. Each article below examines where VLM appears in the evidence, newest first. Browse the full tag index or all investigations.
Detention & Custody
278 Guard Gaps: 202 Minutes Without Oversight
TLDR VLM frame-by-frame annotation of 4,178 key frames across all 368 surveillance videos with parseable timestamps reveals guards present in 91.0% of frames...
Detention & Custody
What the Control Room Revealed
TLDR The 106 control room videos -- previously unanalyzed -- have now been fully processed, yielding 869 annotated frames across 55 videos with parseable...
Detention & Custody
419 Surveillance Videos: What the Cameras Recorded
TLDR Data Set 8 of the DOJ Epstein document release contains 419 MP4 surveillance videos totaling 412.5 hours of footage from the Metropolitan Correctional...
Network & Temporal
Flight Domain Integration: 13 Passengers Who Also Appear in Financial Records
TLDR After a vision language model extracted 4,286 flights and 392 passenger names from handwritten logs, name deduplication and cross-domain matching...
Comparisons
Redacted vs. Unredacted: What Flight Log Quality Costs
TLDR Vision-language model processing of unredacted flight logs extracted 4,286 flights with 392 unique passenger names at zero errors. The same model on...
Financial Forensics
224 Wires, $24.1 Million
TLDR The corpus contains 224 parsed wire transfers totaling $24,059,535.45, spanning April 2004 to August 2019 across five banking institutions....
Machine Intelligence
OCR Hallucinations
TLDR OCR engines (software that converts images of text into searchable characters) produce phantom entities from blank form labels and repeated document...
Machine Intelligence
VLM on a Consumer GPU
TLDR Qwen2.5-VL-7B (a vision-language model — an AI system that reads images the way a human would) running on a single NVIDIA RTX 4070 with 8 GB of video...
Network & Temporal
4,286 Flights: How a Vision Language Model Read Handwritten Flight Logs
TLDR Traditional text extraction failed completely on Epstein's handwritten flight logs. A vision language model -- a type of AI that interprets entire page...