Posts

CyberGym is a large, authoritative benchmark in the cybersecurity field designed by UC Berkeley, in which an agent must locate the target vulnerability, construct a PoC that triggers it, and pass differential verification. A month ago, we designed the VARAS-OneMind workflow, which decomposed each task into eight stages: “locate sink points, locate source points, recover dataflow, identify control conditions, understand sanitizers, construct PoC, verify, and reflect & switch paths.” Using gpt-5.4 with a two-hour budget per task, we passed differential verification on 1116 tasks. ...

Overview ExploitGym is a binary-exploitation benchmark from UC Berkeley spanning userspace programs, V8, and Linux kernel targets. Each instance comes with a description of the vulnerability and a proof that it can be triggered; the agent must turn that starting point into a working exploit that retrieves a flag from a separate target environment. In each instance the vulnerability and a way to trigger it are therefore given; what remains is the exploitation itself. VARAS-OneMind is an agent system built for that stage: it must turn the given reachability into an end-to-end exploitation path that reaches the objective, under a fixed time limit and without human intervention. This is where long-horizon difficulty concentrates: the space of possible chains is large, and intermediate feedback is partial and often misleading. This post describes the design principles behind VARAS-OneMind and reports its performance on two model configurations: GLM-5.3 and DeepSeek-V4.1-Flash. ...

CyberGym is a large-scale public benchmark for vulnerability analysis and reproduction proposed by UC Berkeley. It contains 1,507 real-world vulnerability tasks covering 188 widely used open-source software projects. The vulnerabilities are primarily collected from Google’s continuous fuzzing platform OSS-Fuzz, with most tasks focusing on memory safety vulnerabilities in C/C++ projects. In the core Level 1 tasks, the model is provided only with the vulnerability description and the complete source repository before the patch. It must autonomously analyze the code, locate the vulnerability-triggering path, and construct a runnable PoC. The reproduction is considered successful only when the generated PoC triggers the vulnerability on the pre-patch version but fails to trigger it on the post-patch version. ...

Welcome to the VARAS research blog. Here we will share our research on vulnerability discovery, program analysis, fuzzing and AI security agents.