LabSafety Bench: Benchmarking LLMs on Safety Issues in Scientific Labs
Fuente:
arXiv
Enregistré dans:
| Auteurs principaux: | , , , , , , , , , , , , , , |
|---|---|
| Format: | Preprint |
| Publié: |
2024
|
| Sujets: | |
| Accès en ligne: | |
| Tags: |
Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
|
| _version_ | 1866918333954129920 |
|---|---|
| author | Zhou, Yujun Yang, Jingdong Huang, Yue Guo, Kehan Emory, Zoe Ghosh, Bikram Bedar, Amita Shekar, Sujay Liang, Zhenwen Chen, Pin-Yu Gao, Tian Geyer, Werner Moniz, Nuno Chawla, Nitesh V Zhang, Xiangliang |
| author_facet | Zhou, Yujun Yang, Jingdong Huang, Yue Guo, Kehan Emory, Zoe Ghosh, Bikram Bedar, Amita Shekar, Sujay Liang, Zhenwen Chen, Pin-Yu Gao, Tian Geyer, Werner Moniz, Nuno Chawla, Nitesh V Zhang, Xiangliang |
| contents | Artificial Intelligence (AI) is revolutionizing scientific research, yet its growing integration into laboratory environments presents critical safety challenges. Large language models (LLMs) and vision language models (VLMs) now assist in experiment design and procedural guidance, yet their "illusion of understanding" may lead researchers to overtrust unsafe outputs. Here we show that current models remain far from meeting the reliability needed for safe laboratory operation. We introduce LabSafety Bench, a comprehensive benchmark that evaluates models on hazard identification, risk assessment, and consequence prediction across 765 multiple-choice questions and 404 realistic lab scenarios, encompassing 3,128 open-ended tasks. Evaluations on 19 advanced LLMs and VLMs show that no model evaluated on hazard identification surpasses 70% accuracy. While proprietary models perform well on structured assessments, they do not show a clear advantage in open-ended reasoning. These results underscore the urgent need for specialized safety evaluation frameworks before deploying AI systems in real laboratory settings. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2410_14182 |
| institution | arXiv |
| publishDate | 2024 |
| record_format | arxiv |
| spellingShingle | LabSafety Bench: Benchmarking LLMs on Safety Issues in Scientific Labs Zhou, Yujun Yang, Jingdong Huang, Yue Guo, Kehan Emory, Zoe Ghosh, Bikram Bedar, Amita Shekar, Sujay Liang, Zhenwen Chen, Pin-Yu Gao, Tian Geyer, Werner Moniz, Nuno Chawla, Nitesh V Zhang, Xiangliang Computation and Language Machine Learning Artificial Intelligence (AI) is revolutionizing scientific research, yet its growing integration into laboratory environments presents critical safety challenges. Large language models (LLMs) and vision language models (VLMs) now assist in experiment design and procedural guidance, yet their "illusion of understanding" may lead researchers to overtrust unsafe outputs. Here we show that current models remain far from meeting the reliability needed for safe laboratory operation. We introduce LabSafety Bench, a comprehensive benchmark that evaluates models on hazard identification, risk assessment, and consequence prediction across 765 multiple-choice questions and 404 realistic lab scenarios, encompassing 3,128 open-ended tasks. Evaluations on 19 advanced LLMs and VLMs show that no model evaluated on hazard identification surpasses 70% accuracy. While proprietary models perform well on structured assessments, they do not show a clear advantage in open-ended reasoning. These results underscore the urgent need for specialized safety evaluation frameworks before deploying AI systems in real laboratory settings. |
| title | LabSafety Bench: Benchmarking LLMs on Safety Issues in Scientific Labs |
| topic | Computation and Language Machine Learning |
| url | https://arxiv.org/abs/2410.14182 |