Zhenhua Xu

awesome-llm-copyright-protection. A curated collection of research and techniques for protecting intellectual property of large language models, including watermarking, fingerprinting, and more.

AdaMARP. [ACL2026 Finding]We propose an adaptive multi-agent interaction framework, dubbed AdaMARP, featuring an immersive message format that interleaves [Thought], (Action), <Environment>, and Speech, together with an explicit Scene Manager that governs role-playing through discrete, rationale-annotated actions

EverTracer. [EMNLP2025 MainConference] EverTracer introduces a stealthy and robust gray-box fingerprinting method that verifies large language model ownership by detecting memorized natural-language fingerprints using probability variation signals instead of detectable triggers

CTCC. [EMNLP2025 MainConference] We propose CTCC, a stealthy and robust fingerprinting framework that embeds ownership traces in large language models via semantically structured multi-turn conversation triggers, enabling reliable black-box verification.

LeetCoode-Python. Python

Fingerprint-Vector. We propose Fingerprint Vector, a scalable mechanism that enables post hoc fingerprint transfer across LLM variants, achieving effective, harmless, and robust ownership tracing without direct re-fine-tuning.

ForgetMark. [ICASSP2026]ForgetMark — Stealthy fingerprinting for LLMs via targeted unlearning: build human‑readable key–value sets, embed probabilistic traces, and verify ownership with probability or sematic similarity..

KinGuard. KinGuard, a novel black-box fingerprinting framework that achieves both stealth and robustness by embedding structured kinship knowledge into large language models

VirtualGuard. VirtualGuard: Investigating the use of a continuous virtual token to lock LLMs, revealing a fundamental trade-off between security and performance.

LoRA-FP.