Benchmark RadarAI BENCHMARK PROFILE
AVShift
AVShift evaluates authorship verification across over 150,000 German text pairs spanning three genres and 21 years, with controlled cross-genre, temporal, and AI-era distribution shifts.
- Released
- 2026-08-18
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Existing authorship verification benchmarks study distribution shifts in isolation and focus on English, so AVShift provides a unified German resource to assess robustness under realistic writing-style drifts.
Motivation
Authorship verification (AV) assumes that an author's writing style remains sufficiently stable to distinguish it from that of other writers.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.