MergePrint: Robust Fingerprinting against Merging Large Language Models

no code implementations • 11 Oct 2024 • Shojiro Yamabe, Tsubasa Takahashi, Futa Waseda, Koki Wataoka

As the cost of training large language models (LLMs) rises, protecting their intellectual property has become increasingly critical.

Behavior-Targeted Attack on Reinforcement Learning with Limited Access to Victim's Policy

no code implementations • 6 Jun 2024 • Shojiro Yamabe, Kazuto Fukuchi, Ryoma Senda, Jun Sakuma

In this study, we propose a novel method for manipulating the victim agent in the black-box (i. e., the adversary is allowed to observe the victim's state and action only) and no-box (i. e., the adversary is allowed to observe the victim's state only) setting without requiring environment-specific heuristics.

Imitation Learning reinforcement-learning +1