Benchmark RadarAI BENCHMARK PROFILE
LDU-Bench
LDU-Bench evaluates multimodal LLMs on lithography defect understanding through four tasks: defect triage, morphology recognition, coarse localization, and image-conditioned cause analysis, using task-level metrics and the Lithography Closure Score.
- Released
- 2026-08-04
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Existing industrial anomaly detection benchmarks focus on defect presence, but lithography review requires deeper understanding of morphology, location, and causes. This benchmark aims to quantify those capabilities in a unified platform.
Motivation
Multimodal large language models have demonstrated strong defect recognition capability in industrial anomaly detection.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.