arXiv subjects

Yuxin Bai

Publications and source records attributed to Yuxin Bai.

4 recordsLinked to original sources

arXiv2026

Optimal control of the future via prospective learning with control

Optimal control of the future is the next frontier for AI. Current approaches to this problem are typically rooted in reinforcement learning (RL). RL is mathematically distinct from supervised learning, which has been the main workhorse for the recent achievements in AI. Moreover, RL typically operates in a stationary environment with episodic resets, limiting its utility. Here, we extend supervised learning to address learning to control in non-stationary, reset-free environments. Using this framework, called ''Prospective Learning with Control'' (PLuC), we prove that under certain fairly general assumptions, empirical risk minimization (ERM) asymptotically achieves the Bayes optimal policy. We then consider a specific instance of prospective learning with control: foraging, a canonical task relevant to both natural and artificial agents. We illustrate that modern RL algorithms, which assume stationarity, struggle in these non-stationary reset-free environments. Even with time-aware modifications, they converge orders of magnitude slower than our prospective foraging agents on a simple 1-D foraging benchmark. Code is available at: https://github.com/neurodata/procontrol.

stat.ML↗

arXiv2025

Prospective Learning in Retrospect

In most real-world applications of artificial intelligence, the distributions of the data and the goals of the learners tend to change over time. The Probably Approximately Correct (PAC) learning framework, which underpins most machine learning algorithms, fails to account for dynamic data distributions and evolving objectives, often resulting in suboptimal performance. Prospective learning is a recently introduced mathematical framework that overcomes some of these limitations. We build on this framework to present preliminary results that improve the algorithm and numerical results, and extend prospective learning to sequential decision-making scenarios, specifically foraging. Code is available at: https://github.com/neurodata/prolearn2.

cs.LG↗

arXiv2025

tAIfa: Enhancing Team Effectiveness and Cohesion with AI-Generated Automated Feedback

Providing timely and actionable feedback is crucial for effective collaboration, learning, and coordination within teams. However, many teams face challenges in receiving feedback that aligns with their goals and promotes cohesion. We introduce tAIfa (``Team AI Feedback Assistant''), an AI agent that uses Large Language Models (LLMs) to provide personalized, automated feedback to teams and their members. tAIfa analyzes team interactions, identifies strengths and areas for improvement, and delivers targeted feedback based on communication patterns. We conducted a between-subjects study with 18 teams testing whether using tAIfa impacted their teamwork. Our findings show that tAIfa improved communication and contributions within the teams. This paper contributes to the Human-AI Interaction literature by presenting a computational framework that integrates LLMs to provide automated feedback, introducing tAIfa as a tool to enhance team engagement and cohesion, and providing insights into future AI applications to support team collaboration.

cs.HC↗

arXiv2019

Combining biomarker and self-reported dietary intake data: a review of the state of the art and an exposition of concepts

Classical approaches to assessing dietary intake are associated with measurement error. In an effort to address inherent measurement error in dietary self-reported data there is increased interest in the use of dietary biomarkers as objective measures of intake. Furthermore, there is a growing consensus of the need to combine dietary biomarker data with self-reported data. A review of state of the art techniques employed when combining biomarker and self-reported data is conducted. Two predominant methods, the calibration method and the method of triads, emerge as relevant techniques used when combining biomarker and self-reported data to account for measurement errors in dietary intake assessment. Both methods crucially assume measurement error independence. To expose and understand the performance of these methods in a range of realistic settings, their underpinning statistical concepts are unified and delineated, and thorough simulation studies conducted. Results show that violation of the methods' assumptions negatively impacts resulting inference but that this impact is mitigated when the variation of the biomarker around the true intake is small. Thus there is much scope for the further development of biomarkers and models in tandem to achieve the ultimate goal of accurately assessing dietary intake.

stat.AP↗