Published papers

This paper addresses the system design for virtual network services, particularly focusing on virtual machine (VM) services. The VM service requires users to define a substantial number of specifications, such as CPU resources and latency bounds, as a prerequisite. This process is challenging for nonprofessional users and inefficient even for professional users. Users need to clearly understand the specifications they need and report their requirements to the service manager, which demands a high level of knowledge and experience of a system engineer. To make the system more accessible to users, we propose a framework that enables interaction with the virtual network service through natural language (NL) inputs from the users. The framework employs NL models to interpret user requests in NL format into service specifications to determine the specificationdependent optimal virtual network allocation. We demonstrated the effectiveness of the proposed framework through numerical experiments, which show that the user requests in NL are accurately interpreted and incorporated into the virtual network allocation.

This article proposes a control-based framework for aligning large language models (LLMs) by leveraging a control barrier function (CBF) to ensure user-desirable text generation. The presented framework applies the CBF safety filter to the predicted token generated from the baseline LLM to intervene in the generated text. The safety filter includes two significant advantages: this safety filter is an add-on type, allowing it to be used for alignment purposes without fine-tuning the baseline LLM, and if there is an evaluation model regarding the desired alignment, it can be directly applied to the filter design. The overall text-generation system is implemented with open-source language models, aiming to generate positive text.