Score-Based Diffusion Models for Robust Markowitz Portfolio Optimization; The Clashing Goals for Personality-Emulating Chatbots 87 views

How can we ensure that machine learning models will react well in response to unintended or unforeseen conditions? This principle is demonstrated in both models for constrained optimization and large language models trained to emulate specific personalities. The potential for Score-Based Diffusion models in constrained Markowitz Portfolio Optimization was evaluated in comparison to Primal-Dual and Lagrangian-Dual Learning to Optimize models. Their optimality gaps were compared against the true solutions across baseline data, out of distribution data, and higher-dimensional data than traditionally tested. Score-Based Diffusion models sustained lower optimality gaps across all areas tested. The goals of personality-emulating chatbots are contested, and social groups are competing to shape them in alignment with preferred goals. Some warn that such chatbots may be addictive or otherwise harmful to many users. Nevertheless, chatbot design is substantially controlled by their manufacturers. Some advocacies therefore favor public regulation of personality-emulating chatbots.

BS (Bachelor of Science)

School of Engineering and Applied Science Bachelor of Science in Computer Science Ferdinando Fioretto Peter Norton

English

All rights reserved by the author (no additional license for public reuse)

2025-12-10