# Data-Centric Benchmarking of Exploit Generation in LLMs: Understanding the Impact of Fine-Tuning

arXiv: https://arxiv.org/abs/2606.15123 · HTML: https://mlsanity.com/paper/2606.15123/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 27 of its 122 references that are in the map.
- First posted on arXiv 13 Jun 2026.

## Builds on (most cited first, 27 of 27)

- [Training language models to follow instructions with human feedback](https://mlsanity.com/paper/2203.02155/) (2022)
- [ReAct: Synergizing Reasoning and Acting in Language Models](https://mlsanity.com/paper/2210.03629/) (2022)
- [QLoRA: Efficient Finetuning of Quantized LLMs](https://mlsanity.com/paper/2305.14314/) (2023)
- [The Curious Case of Neural Text Degeneration](https://mlsanity.com/paper/1904.09751/) (2019)
- [Justice or Prejudice? Quantifying Biases in LLM-as-a-Judge](https://mlsanity.com/paper/2410.02736/) (2024)
- [Judging the Judges: A Systematic Study of Position Bias in LLM-as-a-Judge](https://mlsanity.com/paper/2406.07791/) (2024)
- [RLEF: Grounding Code LLMs in Execution Feedback with Reinforcement Learning](https://mlsanity.com/paper/2410.02089/) (2024)
- [LLM Agents can Autonomously Exploit One-day Vulnerabilities](https://mlsanity.com/paper/2404.08144/) (2024)
- [LLM Agents can Autonomously Hack Websites](https://mlsanity.com/paper/2402.06664/) (2024)
- [PentestGPT: An LLM-empowered Automatic Penetration Testing Tool](https://mlsanity.com/paper/2308.06782/) (2023)
- [AutoAttacker: A Large Language Model Guided System to Implement Automatic Cyber-attacks](https://mlsanity.com/paper/2403.01038/) (2024)
- [An Empirical Study of LLM-as-a-Judge for LLM Evaluation: Fine-tuned Judge Model is not a General Substitute for GPT-4](https://mlsanity.com/paper/2403.02839/) (2024)
- [A Theoretical Analysis of the Repetition Problem in Text Generation](https://mlsanity.com/paper/2012.14660/) (2020)
- [The Impact of Initialization on LoRA Finetuning Dynamics](https://mlsanity.com/paper/2406.08447/) (2024)
- [BountyBench: Dollar Impact of AI Agent Attackers and Defenders on Real-World Cybersecurity Systems](https://mlsanity.com/paper/2505.15216/) (2025)
- [Mitigating Overthinking in Large Reasoning Models via Manifold Steering](https://mlsanity.com/paper/2505.22411/) (2025)
- [Analyzing Uncertainty of LLM-as-a-Judge: Interval Evaluations with Conformal Prediction](https://mlsanity.com/paper/2509.18658/) (2025)
- [Wait, Wait, Wait... Why Do Reasoning Models Loop?](https://mlsanity.com/paper/2512.12895/) (2025)
- [Multitask-based Evaluation of Open-Source LLM on Software Vulnerability](https://mlsanity.com/paper/2404.02056/) (2024)
- [D-CIPHER: Dynamic Collaborative Intelligent Multi-Agent System with Planner and Heterogeneous Executors for Offensive Security](https://mlsanity.com/paper/2502.10931/) (2025)
- [TrustJudge: Inconsistencies of LLM-as-a-Judge and How to Alleviate Them](https://mlsanity.com/paper/2509.21117/) (2025)
- [BiasScope: Towards Automated Detection of Bias in LLM-as-a-Judge Evaluation](https://mlsanity.com/paper/2602.09383/) (2026)
- [ELAD: Explanation-Guided Large Language Models Active Distillation](https://mlsanity.com/paper/2402.13098/) (2024)
- [WebDevJudge: Evaluating (M)LLMs as Critiques for Web Development Quality](https://mlsanity.com/paper/2510.18560/) (2025)
- [Good News for Script Kiddies? Evaluating Large Language Models for Automated Exploit Generation](https://mlsanity.com/paper/2505.01065/) (2025)
- [Enhancing AI-based Generation of Software Exploits with Contextual Information](https://mlsanity.com/paper/2408.02402/) (2024)
- [Automatic Techniques to Systematically Discover New Heap Exploitation Primitives](https://mlsanity.com/paper/1903.00503/) (2019)

## Abstract (arXiv)

We study the task of CVE-conditioned exploit generation, where a model drafts proof-of-concept (PoC) exploits given software vulnerability context. We adopt a data-centric approach, constructing a high-quality dataset via multi-stage preprocessing and introducing a scalable evaluation framework with LLM-as-judge and fine-grained rubrics. Under this unified setup, we benchmark 17 large language models across 8 evaluation criteria, providing systematic insights into their zero-shot capabilities. We further show that a compact 8B open-weight model, when fine-tuned on curated data, achieves over 42.5% improvement in exploit quality and rivals some proprietary models when combined with simple test-time rejection strategies. Our results highlight the importance of data quality, structured supervision, and evaluation design for reliable exploit generation, suggesting that these factors can be as critical as model scale in adapting LLMs to cybersecurity tasks.