---
id: 5437
title: "Training State of the Art Vulnerability Discovery Agents through Reinforcement Learning"
source_domain: depthfirst.com
source_url: "https://depthfirst.com/research/dfs-mini1-agent"
published_at: "2026-03-31T00:00:00Z"
tags: ["web3", "smart-contracts", "vulnerability-detection", "ai-agent", "reinforcement-learning", "evmbench", "security-model"]
phying_url: "https://phying.com/item/5437"
lang: en
---

# Training State of the Art Vulnerability Discovery Agents through Reinforcement Learning

## Summary

depthfirst introduces dfs-mini1, a security model co-trained with its harness via reinforcement learning; it reaches Pareto optimality on OpenAI's EVMBench Detect and state-of-the-art pass@8.

## Intelligence Metadata

- **Sources:** depthfirst.com
- **Published:** 2026-03-31T00:00:00Z
- **Original:** https://depthfirst.com/research/dfs-mini1-agent
- **Phying News:** https://phying.com/item/5437