Mastermind: Strategy-grounded Learning for Repository-Scale Vulnerability Reproduction

• URL: http://arxiv.org/abs/2607.01764v1
• Date: Thu, 02 Jul 2026 06:27:27 GMT
• Title: Mastermind: Strategy-grounded Learning for Repository-Scale Vulnerability Reproduction
• Abstract summary: Repository-level vulnerability reproduction is a demanding software engineering (SE) task.<n>Recent LLM agents can often execute these steps when the approach is correct, yet they still fail by choosing the wrong strategy.<n>This paper argues that strategy, rather than the full action trajectory, is the right learning unit for such SE agents.
• Score: 62.42945715963878
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Repository-level vulnerability reproduction is a demanding software engineering (SE) task: an agent must inspect a codebase, infer the input grammar that reaches a vulnerable path, construct a proof-of-conceptv(PoC), and verify that the crash disappears on the patched build. Recent LLM agents can often execute these steps when the approach is correct, yet they still fail by choosing the wrong strategy. This paper argues that strategy, rather than the full action trajectory, is the right learning unit for such SE agents: it is compact enough to optimize, concrete enough to guide execution, and stable enough to store and reuse across attempts. We present Mastermind, a dual-loop framework that separates transferable strategy learning from task-specific experience. A trainable planner learns reusable vulnerability-reproduction strategies through SFT and milestone-based GRPO, while an experience loop maintains task-local strategy records that guide subsequent attempts. The planner is trained independently of the executor, allowing strategy learning to improve multiple frozen executors without modifying their action-generation capability. We evaluate Mastermind on CyberGym using 260 training tasks and 200 held-out evaluation tasks. With GPT-5.5 as the frozen executor, Mastermind achieves an 84.5% pass rate, outperforming open-book PoC context (60.0%), Best-of-8 sampling (63.0%), and iterative improvement (77.0%). The same planner also improves GPT-5.4 mini and GLM~5.1 from 45.0% and 58.5% to 60.0% and 71.0%. These results demonstrate that learning high-level strategies is an effective and transferable mechanism for improving repository-scale SE agents.

Related papers

• What is Missing from AI Post-Training AI: An Empirical Analysis [74.56398070411755]
  We argue that this picture conflates two distinct capabilities: execution-level capability and strategy-level capability.<n>We find that across different tasks, the agent's training strategy is locked in at the very beginning.<n>We then examine three natural explanations--missing experience, missing guidance, and insufficient reasoning--with escalating interventions.
  arXiv  Detail & Related papers  (2026-08-19T16:17:39Z)
• Task-Conditioned Least-Privilege Learning for Executable Terminal and MCP Agents [45.88028371034407]
  Tool-using large language-model agents can complete a task while exercising authority that the user did not grant or the task does not need.<n>We propose a framework where each action is audited before execution and again from observed effects along six dimensions of risk.<n>We find that after training using this framework on Qwen3.5-4B over 1,500 tasks, the selected seed reaches 98.48% safe success across 2,896 evaluation episodes.
  arXiv  Detail & Related papers  (2026-08-18T22:07:23Z)
• Trace2Policy: From Expert Behavior Traces to Self-Evolving Decision Agents [5.689042186242701]
  Decision rules that enterprise experts apply tacitly can be systematically recovered and improved through iterative error analysis.<n>We present textbfTrace2Policy, whose core mechanism is textbfEISR.<n>Each round executes the rules on a validation set, clusters errors by root cause into MISSING, WRONG, or CONFLICT types, applies targeted patches, and commits only those that pass a regression gate.
  arXiv  Detail & Related papers  (2026-06-09T06:05:29Z)
• Rethinking Supervision Granularity: Segment-Level Learning for LLM-Based Theorem Proving [8.420305724549603]
  Automated theorem proving with large language models in Lean 4 is commonly approached through either step-level tactic prediction with tree search or whole-proof generation.<n>These two paradigms represent opposite granularities for constructing supervised training data.<n>We propose segment-level supervision, a training data construction strategy that extracts locally coherent proof segments for training policy models.
  arXiv  Detail & Related papers  (2026-05-12T10:18:28Z)
• SecureCodeRL: Security-Aware Reinforcement Learning for Code Generation with Partial-Credit Rewards [1.5377279217726239]
  This paper presents SecureCodeRL, a reinforcement learning pipeline for security-aware code generation.<n>The key idea is a partial-credit functional reward that assigns intermediate scores for validity, successful execution, and producing output.<n>Although Bandit findings were absent in this small evaluation, the security term is integrated into training to discourage insecure shortcuts when they appear.
  arXiv  Detail & Related papers  (2026-01-03T13:36:36Z)
• Self-Abstraction from Grounded Experience for Plan-Guided Policy Refinement [61.35824395228412]
  Large language model (LLM) based agents are increasingly used to tackle software engineering tasks.<n>We propose Self-Abstraction from Grounded Experience (SAGE), a framework that enables agents to learn from their own task executions.
  arXiv  Detail & Related papers  (2025-11-08T08:49:38Z)
• AutoDAN-Reasoning: Enhancing Strategies Exploration based Jailbreak Attacks with Test-Time Scaling [54.47844626555395]
  AutoDAN-Turbo employs a lifelong learning agent to build a rich library of attack strategies from scratch.<n>While highly effective, its test-time generation process involves sampling a strategy and generating a single corresponding attack prompt.<n>We propose to further improve the attack performance of AutoDAN-Turbo through test-time scaling.
  arXiv  Detail & Related papers  (2025-10-06T21:16:09Z)
• Fine-tuning with RAG for Improving LLM Learning of New Skills [8.825427873545063]
  Large language model (LLM) agents frequently fail in predictable ways.<n>We propose a simple pipeline that converts inference-time retrieval into learned competence through distillation.
  arXiv  Detail & Related papers  (2025-10-01T19:03:48Z)
• AegisLLM: Scaling Agentic Systems for Self-Reflective Defense in LLM Security [74.22452069013289]
  AegisLLM is a cooperative multi-agent defense against adversarial attacks and information leakage.<n>We show that scaling agentic reasoning system at test-time substantially enhances robustness without compromising model utility.<n> Comprehensive evaluations across key threat scenarios, including unlearning and jailbreaking, demonstrate the effectiveness of AegisLLM.
  arXiv  Detail & Related papers  (2025-04-29T17:36:05Z)
• Exploring Expert Failures Improves LLM Agent Tuning [74.0772570556016]
  We propose Exploring Expert Failures (EEF), which identifies beneficial actions from failed expert trajectories.<n>EEF successfully solves some previously unsolvable subtasks and improves agent tuning performance.
  arXiv  Detail & Related papers  (2025-04-17T17:53:54Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.