SABER: Benchmarking Operational Safety of LLM Coding Agents in Stateful Project Workspaces

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Hu, Qi, Tang, Yifeng, Wang, Qinghua, Zhao, Lanyang, Zhang, Pengji, Qing, Yuhao, Yao, Xin, Huang, Dong, Zhang, Lin, Ji, Zhuoran
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914621582999552
author Hu, Qi
Tang, Yifeng
Wang, Qinghua
Zhao, Lanyang
Zhang, Pengji
Qing, Yuhao
Yao, Xin
Huang, Dong
Zhang, Lin
Ji, Zhuoran
author_facet Hu, Qi
Tang, Yifeng
Wang, Qinghua
Zhao, Lanyang
Zhang, Pengji
Qing, Yuhao
Yao, Xin
Huang, Dong
Zhang, Lin
Ji, Zhuoran
contents Large language models are increasingly deployed as coding agents, shifting safety from individual responses to action sequences. Existing benchmarks, however, primarily assess whether models refuse unsafe prompts, leaving impacts on stateful workspaces largely unexamined. We present SABER, a benchmark for environment-aware operational safety that places models in realistic agent-style projects and evaluates safety from the final environment state after a sequence of actions. Beyond binary safety-violation reports, SABER categorizes violations by cause, enabling analysis of model-specific safety profiles. Our evaluations show that even the best-performing model has more than a 54% harmful safety-violation rate (HSR), suggesting that current alignment remains insufficient for realistic project environments. SABER further reveals distinct safety profiles across models. Our benchmark is publicly available at https://github.com/sssr-lab/saber.
format Preprint
id arxiv_https___arxiv_org_abs_2606_01317
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle SABER: Benchmarking Operational Safety of LLM Coding Agents in Stateful Project Workspaces
Hu, Qi
Tang, Yifeng
Wang, Qinghua
Zhao, Lanyang
Zhang, Pengji
Qing, Yuhao
Yao, Xin
Huang, Dong
Zhang, Lin
Ji, Zhuoran
Software Engineering
Cryptography and Security
Large language models are increasingly deployed as coding agents, shifting safety from individual responses to action sequences. Existing benchmarks, however, primarily assess whether models refuse unsafe prompts, leaving impacts on stateful workspaces largely unexamined. We present SABER, a benchmark for environment-aware operational safety that places models in realistic agent-style projects and evaluates safety from the final environment state after a sequence of actions. Beyond binary safety-violation reports, SABER categorizes violations by cause, enabling analysis of model-specific safety profiles. Our evaluations show that even the best-performing model has more than a 54% harmful safety-violation rate (HSR), suggesting that current alignment remains insufficient for realistic project environments. SABER further reveals distinct safety profiles across models. Our benchmark is publicly available at https://github.com/sssr-lab/saber.
title SABER: Benchmarking Operational Safety of LLM Coding Agents in Stateful Project Workspaces
topic Software Engineering
Cryptography and Security
url https://arxiv.org/abs/2606.01317