AI BENCHMARK PROFILE
SopriBench
SopriBench evaluates user-level privacy leakage from social media posts across text, images, and metadata in 50 synthetic profiles with 1,569 images, scoring via the Privacy Exposure Score (PES) that weights value granularity by contextual sensitivity.
- Released
- 2026-06-05
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
This benchmark addresses the gap in evaluating cumulative cross-post privacy leakage, providing a metric that captures exposure severity rather than binary accuracy, aiding in assessing real-world privacy risks from social media data.
Motivation
Public social media posts can reveal private information through weak cues scattered across text, images, or metadata.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.