Peer Pressure: Execution-Verified Co-Evolution of Attacker and Defender Populations for Code Security

Summary

The authors co-train attacker and defender populations of LoRA adapters over one frozen base model, using PRPO (Peer-Relative Policy Optimization), where rewards come from execution in isolated containers rather than static datasets. The resulting 7B diff reviewer reaches 64.8% balanced accuracy on four unseen applications in three untrained languages, 5.5 points above supervised security fine-tuning and 4.4 above a frozen Claude Opus 4.8 reviewer. The 9B repository agent recovers 23 of 172 held-out VulnGym advisories, against 10 after supervised fine-tuning.

Related items

• InfoSafety Mid-Training: Internalizing LLM Safety as a Foundational CapabilitySimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
• InfoConcerns in Benchmarking Jailbreaks: An Empirical Study of Attack Success RatesSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
• InfoBeyond Attack Success Rate: Cost-Aware Metric for Evaluating Iterative LLM JailbreaksSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)
• InfoTraining AIs to Solve CAPTCHAsSimilar attack · Schneier on Security
• InfoProvable Adversarial Detection: Prime Quantization meets Gromov--WassersteinSimilar attack · Arxiv (cs.CR + cs.CL + cs.LG)