Benchmark RadarVisAnomBench
VisAnomBench is a benchmark assembled from public time-series datasets for anomaly detection, augmented with natural-language explanations selected from large vision-language models. It supports fine-tuning a parameter-efficient VLM called VisAnomReasoner for grounded anomaly detection decisions.
- Released
- 2026-05-28
- Readiness
- Paper only
- Primary field
- General AI
Why it matters
Public anomaly detection benchmarks typically lack natural-language rationales, hindering fine-tuning of VLMs for interpretable decisions. VisAnomBench addresses this gap by providing labeled explanations.
Motivation
Recent advances in Vision-Language Models (VLMs) have achieved impressive performance across many tasks, yet prior studies report unsatisfactory performance when applying large language or multimodal models to finding abnormal patterns in sequential data.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.