Benchmark RadarAI BENCHMARK PROFILE
DocPrivacyBench
DocPrivacyBench evaluates susceptibility of document understanding MLLMs to relational privacy leakage when visual evidence is absent or minimal, using KIE tasks on identity documents.
- Released
- 2026-08-13
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Addresses underexplored privacy vulnerabilities in document MLLMs, providing a framework to assess and mitigate leakage of correlated sensitive fields, which is crucial for trustworthy deployment in document processing.
Motivation
While the privacy risks of multimodal large language models (MLLMs) have drawn significant attention, the unique vulnerabilities of domain-specific MLLMs remain largely underexplored.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.