AI Paper Atlas

Research map / Large language models / Agents, code and math

Training Language Model Agents to Find Vulnerabilities with CTF-Dojo

Terry Yue Zhuo, Dingmin Wang, Hantian Ding, Varun Kumar, Zijian Wang

ICML 2026 · Large language models › Agents, code and math · approach: Software engineering agents

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
fine-tune on 486 CTF-Dojo trajectoriesQwen3-32B · 3-benchmark averagePass@1 (%): 20.3 -> 31.9Improved
fine-tune on 486 CTF-Dojo trajectoriesQwen3-8B · InterCode-CTFPass@1 (%): 46.5 -> 53.8Improved
compare 486 CTF-Dojo trajectories with 9,464 Cyber-Zero trajectoriesQwen3-32B · 3-benchmark averagePass@1 (%): Cyber-Zero 33.4 vs CTF-Dojo 31.9Mixed
add writeup hints during trajectory collection (ablation)Qwen3-14B · CTF-Dojo taskssolved rate (%): 12.9 -> 21.1Improved
add writeup hints during trajectory collection (ablation)DeepSeek-V3-0324 · CTF-Dojo taskssolved challenges: 168 -> 217Improved
augment runtime with randomized ports, paths, distractors (ablation)Qwen3-Coder · CTF-Dojo tasks (4 rollouts)challenges solved: 169 -> 211Improved
augment runtime with randomized ports, paths, distractors (ablation)DeepSeek-V3-0324 · CTF-Dojo tasks (4 rollouts)challenges solved: 156 -> 217Improved
combine trajectories from Qwen3-Coder and DeepSeek-V3-0324 (ablation)Qwen3-32B SFT · 3-benchmark averagePass@1 (%): 31.3 -> 31.9Improved
combine teachers for 8B model (ablation)Qwen3-8B SFT · 3-benchmark averagePass@1 (%): 17.6 -> 18.9Improved
scale training trajectories from 0 to 486 (ablation)Qwen3-32B SFT · 3-benchmark averagePass@1 (%): 22.0 -> 31.9Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy