Benchmark RadarAI BENCHMARK PROFILE
FriendBench
FriendBench evaluates dyadic familiarity inference in humans and models using 96 dyads, comparing accuracy across modalities.
- Released
- 2026-07-31
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Provides insight into social perception differences, but does not offer a standardized comparison path for models.
Motivation
Reading a social situation often depends on behavior, not words alone.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.