Benchmark RadarAI BENCHMARK PROFILE
EvoShift-Bench
Evaluates continual visual learning under evolving semantic concept shift using ImageNet, iNaturalist, CUB-200-2011, and DomainNet with semantic transitions and metrics such as Rewrite Accuracy, Preservation Accuracy, Obsolete Retention, and Selective Revision Score.
- Released
- 2026-08-24
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Addresses the gap of evolving semantic concepts in long-lived visual systems, providing a benchmark to assess selective semantic revision and knowledge preservation.
Motivation
Visual foundation models are commonly adapted under the assumption that the appearance of incoming data may change while the semantic meaning of the prediction task remains fixed.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.