MapSteal: Map-based Autonomous Knowledge Stealing Attack Against Large Language Models
We propose MapSteal, a student-independent framework for structured recursive knowledge acquisition.
ProofPaper ↗
Key points
- Knowledge acquisition from black-box large language models (LLMs) offers a practical approach to adapting smaller student models.
- Starting from a small set of initial roots, MapSteal represents acquisition targets using structured states and recursively expands them through teacher-proposed directions.
- Experiments with Qwen3.5 students at 2B, 4B, and 9B scales show competitive downstream performance across six medical and cybersecurity benchmarks.
- These results demonstrate the feasibility of structured, student-independent recursive acquisition while highlighting the distinction between generating acquisition targets and obtaining useful training supervision.
Sources (1)
- [1]MapSteal: Map-based Autonomous Knowledge Stealing Attack Against Large Language ModelsarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 19, 01:07 AM
We propose MapSteal, a student-independent framework for structured recursive knowledge acquisition.
Knowledge acquisition from black-box large language models (LLMs) offers a practical approach to adapting smaller student models.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 19, 2026TokenGate: Adapting LLMs for Fast and Memory-Efficient Prefill with Learned Token Pruning
- Oct 18, 2026Beyond Factual Accuracy: A Benchmark for Evaluating Perspective Alignment in LLMs
- Oct 17, 2026Medical Fine-Tuning Reshapes, Not Reduces, Patient-Facing AI Risk: A Rubric-Based, LLM-Judge Evaluation Framework
- Oct 17, 2026AdamPM: Memory-Efficient LLM Optimization via Moment Pooling
- Oct 17, 2026Otaku — an LLM frontend for creative writing
- Oct 16, 2026ECHO: Replacing the Compliance Proxy with Response Harmfulness in Optimization-Based Jailbreak Attack