Eric Wong

Title: Innovations by Eric Wong in Language Model Security

Introduction

Eric Wong is an innovative inventor based in Philadelphia, PA. He has made significant contributions to the field of artificial intelligence, particularly in the security of large language models (LLMs). His work focuses on developing methods to defend these models against potential vulnerabilities.

Latest Patents

Eric Wong holds a patent for "Methods, systems, and computer readable media for defending large language models (LLMs) against jailbreaking attacks." This patent describes a method for defending an LLM against a jailbreaking attack by receiving an input prompt for generating output from the LLM. The method includes generating N versions of the input prompt, where each version is created by perturbing the original input. The N perturbed input prompts are then provided to the LLM, which generates N responses. The responses are aggregated by estimating a group effect of the perturbed input prompts on the LLM. Finally, one of the N responses is selected based on its agreement with the estimated group effect.

Career Highlights

Eric Wong is affiliated with the University of Pennsylvania, where he continues to advance research in artificial intelligence and machine learning. His work has garnered attention for its practical applications in enhancing the security of AI systems.

Collaborations

Eric collaborates with Alexander Beck Robey, contributing to the innovative research environment at the University of Pennsylvania.

Conclusion

Eric Wong's contributions to the field of language model security highlight the importance of innovation in protecting AI technologies. His patent demonstrates a proactive approach to addressing vulnerabilities in large language models.