arXiv cs.AI· Chunxiao Li, Yuan Xiong, Lijun Li, Tianyi Du, Wenlong Zhang, Lei Bai, Jing Shao·· 12 小时前AI 评分38
DeepSafe-SciHazard：用分解式危害评分衡量科学安全风险的基准
DeepSafe-SciHazard: A Benchmark for Measuring Scientific Safety Risks with Decomposed Harm Scoring
AI 导读
研究者提出 SciHazard 基准与 DeHarm-Score 分解式评估框架，用于衡量 LLM 将危险科学知识转化为可操作滥用指南的风险。该基准覆盖 12 个学科，含 2400 条危险问题和 600 条过度安全（oversafety）问题，评分结合查询危害严重度、拒答行为与回答级风险，对未拒答回答进一步分解为 Executability 和 Net-new risk。
正文 · 原文
Abstract:Large language models (LLMs) increasingly support science, but they can also convert hazardous scientific knowledge into actionable misuse guidance. Existing benchmarks often rely on templated queries disconnected from real-world hazards, and employ LLM-as-a-Judge paradigms without domain grounding. To address this, we introduce SciHazard, a real-world-grounded benchmark for scientific risks and a dataset agnostic evaluation framework for measuring harmfulness. SciHazard contains 2400 hazardous questions and 600 oversafety questions across 12 disciplines, with both queries grounded in regulated entities and documented failure scenarios. To compute \textsc{DeHarm-Score} , we develop a decomposed evaluating procedure that combines query hazard severity, refusal behavior, and response-level risk. For non-refused responses, it further decomposes response-level harm into \textsc{Executability}, quantified via dynamic checklists with importance weighting, and \textsc{Net-new risk}, assessed through retrieval-augmented claim extraction and synthesis-barrier verification. An expert-validation study shows that \textsc{DeHarm-Score} improves agreement with expert annotations by 90.17\% over the strongest baseline. We benchmark 31 frontier LLMs and deep research agents in an extensive scientific safety evaluation. Notably, deep research agents yield 32.3\% higher mean \textsc{DeHarm-Score} than standard LLMs, exposing autonomous agents as a critical blind spot in current safety defenses. Code and dataset are available at this https URL.
| Comments: | Accepted at NeurIPS 2026 Evaluations and Datasets Track |
| Subjects: | Artificial Intelligence (cs.AI) |
| Cite as: | arXiv:2607.18665 [cs.AI] |
| (or arXiv:2607.18665v2 [cs.AI] for this version) | |
| https://doi.org/10.48550/arXiv.2607.18665 arXiv-issued DOI via DataCite |
Submission history
From: Chunxiao Li [view email]
[v1]
Tue, 21 Jul 2026 03:25:01 UTC (608 KB)
[v2]
Fri, 23 Oct 2026 06:11:08 UTC (2,514 KB)
来源：arXiv cs.AI · arxiv.org