Search papers, labs, and topics across Lattice.
This paper introduces VulAgentRL, an agentic reinforcement learning framework designed for interprocedural vulnerability detection, addressing the limitation of existing models that classify functions in isolation. By utilizing a Code Property Graph (CPG) for both querying and verifying evidence, VulAgentRL ensures that the reward system is tied to the quality of evidence supporting vulnerability claims. The results show that VulAgentRL significantly outperforms state-of-the-art baselines, achieving higher accuracy with fewer tool calls, even in challenging scenarios like out-of-distribution data and class imbalance.
Vulnerabilities spanning multiple functions can be detected more accurately with VulAgentRL, which verifies evidence through a novel Code Property Graph approach.
Real-world vulnerabilities often span multiple functions, yet most learning-based detectors classify each function in isolation: on a sample of real CVEs, we find that 71.7% of vulnerable functions require evidence from outside the function to be classified correctly. Agentic reinforcement learning (RL) could close this gap by enabling a model to gather that evidence itself, but it lacks a reliable reward, since a reward defined on the final verdict alone can be obtained without performing any investigation. We propose VulAgentRL, an agentic RL framework for interprocedural vulnerability detection built on a Code Property Graph (CPG). The CPG serves two roles: at inference time the policy queries it for callers, callees, dataflow, and other queries, and at training time the same graph verifies the evidence the policy cites. Because every CPG node carries a persistent integer identifier, this verification is an exact comparison rather than a textual match, so the reward credits verdicts that are supported by evidence. We further initialize the policy by distilling teacher investigations, and show that this warm start is necessary, since RL cannot acquire tool-use behavior it never samples. Under a repository-level split that prevents leakage, VulAgentRL outperforms state-of-the-art baselines, including frontier models, on the strict pair-wise-correct metric while issuing fewer tool calls, and its advantage persists on an out-of-distribution corpus and under class imbalance.
