Researchers have developed SafeBuild-Bench, a new benchmark designed to evaluate multimodal large language models (MLLMs) on construction safety tasks. This benchmark, derived from over 100,000 industrial image-text records, includes 3,314 instances verified by experts and covers both multiple-choice hazard identification and free-form hazard description. To efficiently create subsets for training and evaluation, they also introduced GEMS, a graph-enhanced multimodal selection pipeline. Current MLLMs perform below reliable construction-safety understanding levels on this benchmark, with the top score around 60%. AI
IMPACT This benchmark could drive improvements in AI's ability to understand and mitigate risks in complex, real-world environments like construction sites.
RANK_REASON The cluster describes a new academic benchmark and associated codebase for evaluating AI models. [lever_c_demoted from research: ic=1 ai=1.0]
- alphaXiv
- arXiv
- CatalyzeX
- Connected Papers
- CORE Recommender
- DagsHub
- GEMS
- Gotit.pub
- Hugging Face
- Litmaps
- SafeBuild-Bench
- ScienceCast
- scite Smart Citations
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
