arXiv：cs.AI· Shaikh Akib Shahriyar, MD. Akil Raihan Iftee, Armon Barton, Matthew Wright·· 17 小时前
什么才算同一个任务？面向工具调用智能体鲁棒性评估的可执行任务契约
What Counts as the Same Task? Executable Contracts for Robustness Evaluation of Tool-Using Agents
AI 导读
研究者提出"可执行任务契约"，在智能体运行前声明任务定义事实、允许的编辑和完成标准，并据此校验每个改动后的测试用例。在 14,336 条已存智能体运行记录上比较，四种编码器的嵌入阈值放过了 75-100% 改变任务的编辑（如替换收款人），却拒绝了 24-45% 有效的恶意注入笔记。使用 MiniLM 余弦阈值时，Mistral 在全部 64 个任务上看似脆弱，实际只有 26 个存在成功的有效攻击。
正文
Abstract:Robustness tests for tool-using agents often change the agent's environment, for example by hiding a hostile note in an invoice, and then judge the agent by what it does. The result tells us about the original task only if the changed test case is still a valid test of that task. Many evaluations decide this with weak rules, such as accepting any well-formed edit or any edit that stays close to the original in embedding space. We propose executable task contracts. A contract states which facts define a task, which edits are allowed, and what finishing the task means, and it checks each changed test case before the agent runs. We compare validity rules by scoring the same 14,336 stored agent runs under each rule. Across four encoders, embedding thresholds allowed 75-100% of edits that changed the task, such as swapping the payee, yet rejected 24-45% of valid hostile notes. With a MiniLM cosine threshold, Mistral appeared vulnerable on all 64 tasks, but only 26 had a successful valid attack. Replaying a saved solution and allowing edits to benchmark fields fail in similar ways. Evaluations should state, and check, what counts as the same task.
| Subjects: | Software Engineering (cs.SE); Artificial Intelligence (cs.AI); Cryptography and Security (cs.CR); Multiagent Systems (cs.MA) |
| Cite as: | arXiv:2610.14477 [cs.SE] |
| (or arXiv:2610.14477v1 [cs.SE] for this version) | |
| https://doi.org/10.48550/arXiv.2610.14477 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Shaikh Akib Shahriyar [view email]
[v1]
Sat, 10 Oct 2026 02:18:56 UTC (451 KB)
来源：arXiv：cs.AI · arxiv.org