MapSteal helps smaller AI models learn from large language models autonomously
MapSteal: Map-based Autonomous Knowledge Stealing Attack Against Large Language Models
Computation and LanguageArtificial Intelligence
Summary
Getting small AI models to learn from big ones often needs careful questions or feedback, which can be tricky. The authors propose MapSteal, a new way for smaller models to gather knowledge on their own by exploring topics step-by-step without needing feedback from the smaller models during learning. Their experiments show this method works well, especially in medical and cybersecurity tests. This approach lets smaller AI systems learn from big ones more independently and with better organization.
What this means in practice
- •For medical data teams: Build smaller diagnostic AI tools that learn medical knowledge from large language models without manual question design.
- •For cybersecurity engineers: Develop lightweight models with improved cybersecurity threat detection based on knowledge autonomously extracted from large language models.
Authors
Zhengao Li, Chuxu Zhang, Yushun Dong
Abstract
Knowledge acquisition from black-box large language models (LLMs) offers a practical approach to adapting smaller student models. However, many existing methods rely on predefined queries or student feedback to guide acquisition, leaving autonomous expansion of knowledge targets an important challenge. We propose \textbf{MapSteal}, a student-independent framework for structured recursive knowledge acquisition. Starting from a small set of initial roots, MapSteal represents acquisition targets using structured states and recursively expands them through teacher-proposed directions. Question realization, proposal admission, and deduplication constrain the acquisition process, while root-aware scheduling organizes exploration under a fixed QA budget. The resulting question--answer pairs are used to adapt student models without requiring student feedback during acquisition. Experiments with Qwen3.5 students at 2B, 4B, and 9B scales show competitive downstream performance across six medical and cybersecurity benchmarks. On the 4B medical evaluation, Full MapSteal improves average accuracy by 1.17 percentage points over a simplified recursive acquisition variant. Cross-domain evaluations show that gains depend on the domain and evaluation panel. These results demonstrate the feasibility of structured, student-independent recursive acquisition while highlighting the distinction between generating acquisition targets and obtaining useful training supervision. Our code is available at \href{https://github.com/LabRAI/Mapsteal} {https://github.com/LabRAI/Mapsteal}.