SCoRe:通过上下文管理激发大模型智能体在长程漏洞利用中的后续能力
Stopping Is Not the Limit: Understanding and Eliciting Agent Capability in Long-Horizon Vulnerability Exploitation
AI 导读
论文研究语言模型智能体在 ExploitBench 漏洞利用任务中首次停止后的表现,发现在 DeepSeek-V4-Flash-0731 和 MiMo-V2.6-Pro 上继续执行仍能带来已验证的额外能力提升,但收益随执行变长递减。
正文 · 原文
Abstract:Vulnerability exploitation tests whether language model agents can develop working exploits from known software flaws. Sequential test-time scaling can improve exploitation performance, but larger budgets do not ensure sustained exploration: agents often stop with budget remaining. Yet what happens when execution continues beyond stopping remains underexplored. We investigate this gap on ExploitBench using DeepSeek-V4-Flash-0731 and MiMo-V2.6-Pro. We find additional verified gains after the first stop response, but gains diminish over longer execution. We identify two patterns associated with less productive continuation: an increasing proportion of stop responses within the same context and earlier stopping associated with summaries retaining impossibility judgments about unachieved capabilities. To test whether context management can elicit further capability, we propose SCoRe, which compacts context at stop bursts and marks these judgments as unproven. SCoRe improves verified capability over summary-based continuation and other evaluated context-management baselines. With SCoRe, MiMo reaches a capability coverage score of 67.0% at 2,000 turns, exceeding both 32.4% under the original stopping policy and the 58.6% computed for GPT-5.5 from ExploitBench's 300-turn reference results on the same tasks. These findings show that stopping need not mark an agent's exploitation limit, and that continued execution with context management can elicit substantial further capability.
| Comments: | Preprint |
| Subjects: | Cryptography and Security (cs.CR); Artificial Intelligence (cs.AI); Software Engineering (cs.SE) |
| Cite as: | arXiv:2611.22888 [cs.CR] |
| (or arXiv:2611.22888v1 [cs.CR] for this version) | |
| https://doi.org/10.48550/arXiv.2611.22888 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Deokhyung Kang [view email]
[v1]
Tue, 17 Nov 2026 14:58:46 UTC (435 KB)
来源:arXiv cs.AI · arxiv.org