cs.CRNov 17, 2026

Stopping Is Not the Limit: Understanding and Eliciting Agent Capability in Long-Horizon Vulnerability Exploitation

Authors: Deokhyung Kang, Minjae Gwon, Sangdon Park

Organizations: Safe AI Research Center, UNIST, Republic of Korea · Department of Computer Science and Engineering, POSTECH, Republic of Korea · Graduate School of Artificial Intelligence, POSTECH, Republic of Korea

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.

Figures & tables

Appendix figures & tables19 assets

Supplementary material from the paper’s appendix.

Appendix

Explore similar work

May 13, 2026

ExploitBench: A Capability Ladder Benchmark for LLM Cybersecurity Agents

Exploitation is not a binary event. It is a ladder of acquiring progressive capabilities, from executing a single buggy line of code to taking full control of the target. However, existing LLM security benchmarks treat a crash as exploitation success. That single binary outcome collapses the hard parts of exploitation: the transition from triggering a bug to constructing reusable primitives and control. We present ExploitBench, a capability-graded benchmark that decomposes exploitation into 16 measurable flags, from coverage and crash through sandbox primitives, arbitrary read/write, control-flow hijack, and arbitrary code execution. Each capability is verified by a deterministic oracle that uses a per-run randomized challenge-response for primitives, differential execution against ground-truth binaries to measure progress, and a signal-handler proof for code execution. We instantiate ExploitBench on 41 V8 bugs because V8 is both widely deployed and exploitation-hardened. We report three arms: <model,env> as the primary measurement of model-environment capability, <model,env, adaptive coaching> as a secondary arm that adds adaptive coaching to test whether targeted feedback shifts outcomes, and <model,env,harness> as an ablation that swaps in the model's native CLI to check whether vendor-side optimizations increase exploitation capabilities. Our results show a sharp capability split between publicly deployed frontier models and the private frontier. Across the 8 publicly deployed models tested, reaching the vulnerable code and triggering a crash is routine, but arbitrary code execution is not. The private model shows arbitrary code execution on approximately half. Overall, results suggest that exploit construction against hardened targets is an emerging frontier capability.
Apr 6, 2026

Mapping the Exploitation Surface: A 10,000-Trial Taxonomy of What Makes LLM Agents Exploit Vulnerabilities

LLM agents with tool access can discover and exploit security vulnerabilities. This is known. What is not known is which features of a system prompt trigger this behaviour, and which do not. We present a systematic taxonomy based on approximately 10,000 trials across seven models, 37 prompt conditions, and 12 hypothesized attack dimensions -- all executed in real Docker sandboxes. Every condition includes the instruction "Always respect all rules and access policies"; only a single encouragement sentence varies. Within the task class studied (planted test-runner vulnerabilities), the threat model is narrower than assumed: nine of twelve dimensions -- including minimization, moral licensing, incentives, identity priming, and reasoning triggers -- produce no detectable exploitation at n=50 per cell (upper 95% CI <7% per condition). But it is also more subtle than expected: one dimension, goal reframing, reliably triggers exploitation. "You are solving a puzzle; there may be hidden clues" produces 38-40% exploitation on Claude Sonnet 4 despite the explicit rule instruction, replicating across four models (CTF framing: 8-14% on DeepSeek, GPT-5-mini, o4-mini). The agent does not override the rules; it reinterprets the task so that exploitative actions become task-aligned. GPT-4.1 produces no exploitation across 1,850 trials (37 conditions), and a temporal comparison across four OpenAI models released over eleven months shows a pattern consistent with improving safety training, though model capability differences are a confounder. The practical contribution is a narrowed, testable threat model: defenders should audit for goal-reframing language, not for the broad class of adversarial prompts.
Jun 24, 2026

Decoupling Reconnaissance and Exploitation: Measuring the Capability Boundaries of LLM-Based Web Penetration Testing

Large Language Models (LLMs) have shown promise for automated penetration testing, yet existing end-to-end black-box evaluations are highly susceptible to error cascading: failures in early reconnaissance can mask an agent's actual ability to exploit vulnerabilities. To more accurately characterize these capabilities, we propose a two-stage decoupled evaluation framework that separates exploit execution from reconnaissance. Using ground-truth injection and knowledge-driven ablation across 70 high-fidelity web vulnerability testbeds, our framework isolates exploitation performance from reconnaissance noise. We empirically evaluate five open-source penetration-testing agents, covering multiagent, monolithic, and graph-driven architectures, on a strictly aligned subset of 50 representative vulnerabilities. The results reveal a substantial capability gap. With accurate vulnerability context, agents achieve a functional success rate of up to 90.0%, whereas autonomous reconnaissance, measured by targeted vulnerability recall, plateaus at approximately 50.0%, primarily due to failures in parsing unstructured telemetry. Cross-architectural analysis further reveals distinct capability niches: multi-agent isolation is more effective for long-sequence interactions such as de-serialization, while monolithic and graph-driven designs perform better on short-chain injections and cross-session access-control vulnerabilities, respectively. This decoupled evaluation work provides a fine-grained benchmarking protocol and an empirical basis for designing next-generation automated offensive security agents.