Engineering topics
Publications and source records attributed to Cooley, Rafer Scott.
Analysis on the effectiveness of instruction tuned embedding models for the reduction of Artificial Intelligence hallucinations compared to the standard embedding models.
Large Language Models (LLMs), such as ChatGPT, tend to “hallucinate”, meaning they confidently generate false information. Retrieval Augmented Generation (RAG) attempts to diminish hallucination by providing context to the LLM from data stores (indexes) containing relevant information. The LLM uses this context to formulate its response. RAG systems can still suffer from hallucination because of bad embeddings or ineffective routing. For example, a router will often return context from an irrelevant index, resulting in a hallucinated answer. In this study, we aim to minimize the frequency of routing hallucinations by optimizing Index Summary Routing.
This is a poster submission, so we do not have a formal abstract; however, we have included a couple of the main points of the poster. Motivation Current ICS/IIoT reference architectures are ill-suited to continue to meet industry demands. Architectures such as the Purdue Model and ISA-95 split cyber-physical systems into distinct levels; however, business needs and cybersecurity requirements constantly require cross-cutting solutions. Further, newly defined IIoT architectures are better suited for Industry 4.0 systems, but they leave out the feedback seen in Purdue and ISA-95. This underscores a need for new architectures, without which IT and OT boundaries will continue to grow. Secure embedded intelligence (SEI) should drive the creation of these new architectures. Secure Embedded Intelligence Advances in sensing, computation, and automated analysis have opened new possibilities for increased security, performance, and cost savings. SEI enables advanced technology to be utilized to drive new reference architectures to achieve these possibilities while integrating trust and risk.