论文检索

Matteo Baldoni, Cristina Baroglio, Monica Bucciarelli, Sara Capecchi, Leonardo Castellani, Elena Gandolfi, Francesco Ianì, Elisa Marengo, Roberto Micalizio

Theory of mind refers to the attribution of mental states that humans ascribe to other humans or objects (such as computer-based systems). Recently, the attribution of mental states has been investigated toward Artificial Intelligence (AI) as a basic manner to capture people's engagement toward it, and people's perception about AI social skills and AI capabilities. In line with this idea, mental state attribution can be used as an indirect measure of students' understanding of AI functioning, and in particular of the kind of interactions students may have with AI systems. Too often is the case of people using generative AI systems in ways that exceed their actual ways of functioning. In our study, children of age in the range 9-12 were involved in one-shot unplugged activities concerning data and models. The unplugged activities were not aimed at teaching the theory of Machine Learning, but rather they were designed so as to provide awareness on some basic mechanisms and help developing a correct use of tools that are becoming more and more present in everyday life. This paper introduces the activities and reports the results that were achieved.

Budhitama Subagdja, Shanthoshigaa D, Ah-Hwee Tan, Iris Rawtaer

This paper introduces a novel system for in-home cognitive health assessment using ambient sensors and a machine learning technology that can robustly detect mild cognitive impairment (MCI) despite limited available data. The learned model can explain the aspects of individuals' daily lives led to the prediction, while reliably predicting MCI, providing more insights to healthcare workers for further clinical interventions. We developed the robust transparent machine learning model, based on fusion adaptive resonance theory (Fusion ART) neural network to learn individuals' daily patterns of activity from continuous sensor data in terms of a suite of digital biomarkers reflecting four key domains: physical, daily activity, cognitive engagement, and sleep patterns. Based on a longitudinal study of over one hundred participants, deployed with non-intrusive sensors in their homes to undergo parallel clinical evaluation across a period of five years, our model successfully identified individuals with MCI, achieving high predictive accuracy regardless the noisy and sparse availability of data. As a transparent neural network, the learned model can also be interpreted as classification rules to distinguish MCI from normal cognition (NC) cases based on the digital biomarkers. These results demonstrate that passively collected, sensor-derived digital biomarkers can be leveraged to indicate cognitive status and potentially providing clinically meaningful insights on the impairment conditions. We also discuss the practical challenges and lessons learned from this real-world deployment to inform future large-scale implementations of such AI-driven health monitoring systems.

Wijnand Van Woerkom, Davide Grossi, Henry Prakken, Bart Verheij

The widespread application of uninterpretable machine learning systems for sensitive purposes has spurred research into elucidating the decision-making process of these systems. These efforts have their background in many different disciplines, one of which is the field of AI & law. In particular, recent works have observed that machine learning training data can be interpreted as legal cases. Under this interpretation, the formalism developed to study case law, called the theory of precedential constraint, can be used to analyze the way in which machine learning systems draw on training data—or should draw on them—to make decisions. In the present work, we advance the theory underlying these explanation methods, by relating it to order theory and logic. This allows us to write a software implementation of the theory that can be used to compute with the definitions and give automatic proofs of the properties of the model. We use this implementation to evaluate the model on a series of datasets. Through this analysis, we characterize the types of datasets that are more, or less, suitable to be described by the theory.

Mourad El Hamri, Younès Bennani, Issam Falih

Domain adaptation arises as an important problem in statistical learning theory, arising when the data-generating processes differ between the training and test samples, respectively called source and target domains. Recent theoretical advances have demonstrated that the success of domain adaptation algorithms heavily relies on their ability to minimize the divergence between the probability distributions of the source and target domains. However, minimizing this divergence cannot be achieved independently of other key ingredients, such as the source risk or the combined error of the ideal joint hypothesis. The trade-off between these terms is often ensured through algorithmic solutions that remain implicit and are not directly reflected by the theoretical guarantees. To get to the bottom of this issue, we propose in this paper a new theoretical framework for domain adaptation through hierarchical optimal transport. This framework provides more explicit generalization bounds and enables us to consider the natural hierarchical organization of samples in both domains into structures, i.e. classes or clusters. Additionally, we provide a new divergence measure between the source and target domains called Hierarchical Wasserstein distance that indicates under mild assumptions, which structures need to be aligned to achieve successful adaptation.

Weitong Zhang

In this New Faculty Highlights, I begin with the reward free exploration that learns broad state and skill coverage with intrinsic rewards and remains robust under misspecification during efficient finetuning; guided generation methods that preserve the prior policy and mitigate reward hacking; and AI for science and healthcare, including practical RL for autonomous laboratories and automatic diagnosis. Building on impacts evidenced by publications, adoption, and awards. My future work will pursue imitation learning and contextual multi task RL that connect behavioral cloning with interactive policies without explicit reward design; personalized and multi-tasked offline to online adaptation with in-context demonstrations. In parallel, I am broadening the impact of AI for science and healthcare through existing collaborations. I will close with a talk that surveys these results and outlines an agenda for reinforcement learning without explicit rewards.

Honghao Wei

Safe reinforcement learning (RL) has emerged as a key paradigm for deploying AI in high-stakes domains such as autonomous driving, robotics, healthcare, and recommender systems. By embedding constraints into the learning process, safe RL enables agents to optimize performance while satisfying critical requirements, including collision avoidance, resource limits, and system reliability. Such guarantees are indispensable for real-world AI, where failures can cause physical harm, economic loss, or loss of trust. At the same time, demand for trustworthy AI continues to grow as machine learning is increasingly deployed in human-centered applications. This makes it essential to design RL algorithms that are not only efficient but also reliable, robust, and aligned with societal needs.

Jasper C.H. Lee

Given society's increasing reliance on data, its collection and processing into useful information is a technical problem of growing focus, and perhaps paradoxically, a critical bottleneck in many data science and machine learning applications. Yet, even for the most basic statistical problems such as mean estimation, there is a theory-practice divide. Conventional methods like the sample mean, while supported by theoretical results under strong assumptions, are often brittle in the presence of extreme data. Practitioners thus often use ad-hoc and unprincipled "outlier removal" heuristics, but which can lead to wrong conclusions (e.g. Milikan's underestimation of the electron charge). In this talk, I will describe my work that essentially resolves the fundamental 1-d mean estimation problem. I will show the construction of a statistically-optimal and computationally-efficient 1-dimensional mean estimator, whose estimation error is optimal even in the leading multiplicative constant, under bare minimum distributional assumptions (FOCS 2021). Furthermore, I will discuss its various robustness properties (ICML 2025 Oral), in particular highlighting robustness to adversarial sample corruption.

Sridhar Mahadevan

We propose a new theoretical foundation for artificial intelligence (AI) and machine learning (ML), building on ideas in pure mathematics relating to categories and functors. This paper builds on our AAAI 2025 tutorial Thinking with Functors: Category Theory for A(G)I, which provides background material. In addition, our recent papers on intuitionistic j-do calculus in Topos Causal Models} and GAIA: Categorical Foundations of Generative AI, illustrate how to generalize well-known formalisms in AI, such as causal inference and deep learning, to a category-theoretic setting.

Dorien Herremans, Abhinaba Roy

Recent advances in generative AI for music have achieved remarkable fidelity and stylistic diversity, yet these systems often fail to align with nuanced human preferences due to the specific loss functions they use. This paper advocates for the systematic application of preference alignment techniques to music generation, addressing the fundamental gap between computational optimization and human musical appreciation. Drawing on recent breakthroughs including MusicRL's large-scale preference learning, multi-preference alignment frameworks like diffusion-based preference optimization in DiffRhythm+, and inference-time optimization techniques like Text2midi-InferAlign, we discuss how these techniques can address music's unique challenges: temporal coherence, harmonic consistency, and subjective quality assessment. We identify key research challenges including scalability to long-form compositions, reliability amongst others in preference modelling. Looking forward, we envision preference-aligned music generation enabling transformative applications in interactive composition tools and personalized music services. This work calls for sustained interdisciplinary research combining advances in machine learning, music-theory to create music AI systems that truly serve human creative and experiential needs.

Chenyu Zhang, Xiaohang Luo

Formative feedback is widely recognized as one of the most effective drivers of student learning, yet it remains difficult to implement equitably at scale. In large or low-resource courses, instructors often lack the time, staffing, and bandwidth required to review and respond to every student reflection, creating gaps in support precisely where learners would benefit most. This paper presents a theory-grounded system that uses five coordinated role-based LLM agents (Evaluator, Equity Monitor, Metacognitive Coach, Aggregator, and Reflexion Reviewer) to score learner reflections with a shared rubric and to generate short, bias-aware, learner-facing comments. The agents first produce structured rubric scores, then check for potentially biased or exclusionary language, add metacognitive prompts that invite students to think about their own thinking, and finally compose a concise feedback message of at most 120 words. The system includes simple fairness checks that compare scoring error across lower and higher scoring learners, enabling instructors to monitor and bound disparities in accuracy. We evaluate the pipeline in a 12-session AI literacy program with adult learners. In this setting, the system produces rubric scores that approach expert-level agreement, and trained graders rate the AI-generated comments as helpful, empathetic, and well aligned with instructional goals. Taken together, these results show that multi-agent LLM systems can deliver equitable, high-quality formative feedback at a scale and speed that would be impossible for human graders alone. The approach demonstrates how structured agent roles, fairness checks, and learning-science principles can work together to support instructors while preserving pedagogical intent. More broadly, the work points toward a future where feedback-rich learning becomes feasible for any course size or context, advancing long-standing goals of equity, access, and instructional capacity in education.

Qingqing Liang, Chunyang Wang, Peiwei Xia, Yanan Zeng, Xin Liu, Xuesong Lu, Aoying Zhou

Educational recommendation systems have been a fundamental component for alleviating learning disorientation in self-paced learning. While existing studies mainly leverage cognitive theories to guide learning motivation modeling, they critically overlook the role of social influences. Through empirical analysis, we identify social homophily as an additional driver of learning behaviors, i.e., learners tend to adopt resources validated by their social cohort. However, two challenges impede effective social homophily modeling: (1) the absence and sparsity of predefined social relations in online education, and (2) the deep entanglement of social homophily with cognitive homophily in behavioral data. To tackle these challenges, we propose a graph-based framework EdGCL that explicitly disentangles social homophily and cognitive homophily. EdGCL infers implicit social relations from learners' social behaviors and encodes them via a graph transformer, generating social-view representations. Simultaneously, it constructs a heterogeneous learning graph to model cognitive homophily, which is enhanced by a type-aware aggregator and cognitive diagnosis loss. To ensure the semantic distinctiveness of dual-view homophily modeling, a cross-view contrastive disentanglement mechanism is designed to pull intra-view representations closer while pushing inter-view representations away. Evaluation on two real-world educational datasets demonstrates the superior recommendation performance of EdGCL, highlighting the necessity of dual homophily modeling for understanding the motivations behind learning behaviors.

Calvin Isley, Joshua Gilbert, Evangelos Kassos, Michaela Kocher, Allen Nie, Emma Brunskill, Ben Domingue, Jake Hofman, Joscha Legewie, Teddy Svoronos 等

While large language models (LLMs) challenge conventional methods of teaching and learning, they present an exciting opportunity to improve efficiency and scale high-quality instruction. One promising application is the generation of customized exams, tailored to specific course content. There has been significant recent excitement on automatically generating questions using artificial intelligence, but also comparatively little work evaluating the psychometric quality of these items in real-world educational settings. Filling this gap is an important step toward understanding generative AI's role in effective test design. In this study, we introduce and evaluate an iterative refinement strategy for question generation, repeatedly producing, assessing, and improving questions through cycles of LLM-generated critique and revision. We evaluate the quality of these AI-generated questions in a large-scale field study involving 91 classes---covering computer science, mathematics, chemistry, and more---in dozens of colleges across the United States, comprising nearly 1700 students. Our analysis, based on item response theory (IRT), suggests that for students in our sample the AI-generated questions performed comparably to expert-created questions designed for standardized exams. Our results illustrate the power of AI to make high-quality assessments more readily available, benefiting both teachers and students.

Evan Dong, Nikhil Garg, Sarah Dean

Machine learning models are often used to make predictions about the outcomes of applications to selective programs. Many prospective school or college applicants turn to machine learning models to predict whether they will be admitted to a program, and employers may use algorithmic tools to filter out resumes predicted to have a low probability of being hired when offering interviews for a job opening. However, such decision processes differ substantially from the conventional machine learning setting: decisions are not independent across applicants. Whether a student is admitted depends on the other applicants who apply because admissions decisions are capacity-constrained. We formalize how the nature of admission decisions results in a data-generating process which is incompatible with traditional machine learning assumptions. We characterize how selection functions properties affect the difficulty of generalization to applicant pool distribution shifts, introducing two concepts: stability, which measures how many existing decisions can change when a single new applicant is introduced; and variability, which measures the number of unique students whose decisions can change. We demonstrate our theory on admissions data from the New York City high school matching system, showing that machine learning performance degrades as the applicant pool increasingly differs from the training data. Furthermore, there are larger performance drops for schools using decision rules that are less stable and more variable. Our work raises questions about the reliability of predicting individual admissions probabilities.

Ankit Bhardwaj, Rohail Asim, Sachin Kumar Chauhan, Yasir Zaki, Lakshmi Subramanian

Free-flow road networks, such as suburban highways, are increasingly experiencing traffic congestion due to growing commuter inflow and limited infrastructure. Traditional control mechanisms—traffic signals or local heuristics—are ineffective or infeasible in these high-speed, signal-free environments. We introduce self-regulating cars, a reinforcement learning-based traffic control protocol that dynamically modulates vehicle speeds to optimize throughput and prevent congestion, without requiring new physical infrastructure. Our approach integrates classical traffic flow theory, gap acceptance models, and microscopic simulation into a physics-informed RL framework. By abstracting roads into super-segments, the agent captures emergent flow dynamics and learns robust speed modulation policies from instantaneous traffic observations. Evaluated in the high-fidelity PTV Vissim simulator on a real-world highway network, our method improves total throughput by 5%, reduces average delay by 13%, and decreases total stops by 3% compared to the no-control setting. It also achieves smoother, congestion-resistant flow while generalizing across varied traffic patterns—demonstrating its potential for scalable, ML-driven traffic management.

Ali Falahati, Mohammad Mohammadi Amiri, Kate Larson, Lukasz Golab

In self-consuming generative models that train on their own outputs, alignment with user preferences becomes a recursive rather than one-time process. In this paper, we provide the first formal foundation for analyzing the long-term effects of such recursive retraining on alignment. Under a two-stage curation mechanism based on the Bradley–Terry (BT) model, we model alignment as an interaction between two factions: the Model Owner, who filters which outputs should be learned by the model, and the Public User, who determines which outputs are ultimately shared and retained through interactions with the model. Our analysis reveals three structural convergence regimes: consensus collapse, compromise on shared optima, and asymmetric refinement, depending on the degree of preference alignment. We prove a fundamental impossibility theorem: no recursive BT-based curation mechanism can simultaneously preserve diversity, ensure symmetric influence, and eliminate dependence on initialization. Framing the process as dynamic social choice, we show that alignment is not a static goal but an evolving equilibrium shaped by power asymmetries and path dependence.

Maxim Divilkovskiy, Alexander Gasnikov

We study non-smooth stochastic decentralized optimization problems over time-varying networks, where objective functions are distributed across nodes and network connections may intermittently appear or break. Specifically, we consider two settings: (i) stochastic non-smooth (strongly) convex optimization, and (ii) stochastic non-smooth (strongly) convex–(strongly) concave saddle point optimization. Convex problems of this type commonly arise in deep neural network training, while saddle point problems are central to machine learning tasks such as the training of generative adversarial networks (GANs). Prior works have primarily focused on the smooth setting, or time-invariant network scenarios. We extend the existing theory to the more general non-smooth and stochastic setting over time-varying networks and saddle point problems. Our analysis establishes upper bounds on both the number of stochastic oracle calls and communication rounds, matching lower bounds for both convex and saddle point optimization problems.

Masataro Asai, Stephen Wissow

Despite being successful in board games and reinforcement learning (RL), Monte Carlo Tree Search (MCTS) combined with Multi Armed Bandit (MAB) has seen limited success in domain-independent classical planning until recently. Previous work (Wissow and Asai, 2024) showed that UCB1, designed for bounded rewards, does not perform well as applied to cost-to-go estimates in classical planning, because cost-to-go estimates are unbounded, and showed improved performance using a Gaussian reward MAB instead. This paper further sharpens our understanding of ideal bandits for planning tasks. Existing work has two issues: first, Gaussian MABs under-specify the support of cost-to-go estimates as (-∞, ∞), which we can narrow down. Second, Full Bellman backup (Schulte and Keller, 2014) that backpropagates sample max/min lacks theoretical justification. We use Peaks-Over-Threashold Extreme Value Theory to resolve both issues at once, propose a new bandit algorithm (UCB1-Uniform). We formally prove its regret bound and empirically demonstrate its performance in classical planning.

Chenyu Zhang, Lanjun Wang, Yiwen Ma, Wenhui Li, Guoqing Jin, Anan Liu

Text-to-Image (T2I) models typically deploy safety mechanisms to prevent the generation of sensitive images. Unfortunately, recent jailbreaking attack methods manually design instructions for the LLM to generate adversarial prompts, which effectively exposing safety vulnerabilities of T2I models. However, existing methods have two limitations: 1) relying on manually exhaustive strategies for designing adversarial prompts, lacking a unified framework, and 2) requiring numerous queries to achieve a successful attack, limiting their practical applicability. To address this issue, we propose Reason2Attack~(R2A), which aims to enhance the effectiveness and efficiency of the LLM in jailbreaking attacks. Specifically, we first use Frame Semantics theory to systematize existing manually crafted strategies and propose a unified generation framework to generate CoT adversarial prompts step by step. Following this, we propose a two-stage LLM reasoning training framework guided by the attack process. In the first stage, the LLM is fine-tuned with CoT examples generated by the unified generation framework to internalize the adversarial prompt generation process grounded in Frame Semantics. In the second stage, we incorporate the jailbreaking task into the LLM's reinforcement learning process, guided by the proposed attack process reward function that balances prompt stealthiness, effectiveness, and length, enabling the LLM to understand T2I models and safety mechanisms. Extensive experiments on various T2I models with safety mechanisms, and commercial T2I models, show the superiority and practicality of R2A.

Chenghao Xu, Jiexi Yan, Muli Yang, Fen Fang, Huilin Chen, Cheng Deng

Large Language Models (LLMs) are prone to generating incorrect or outdated information, thereby necessitating efficient and precise mechanisms for knowledge updates. Existing knowledge editing approaches, however, often encounter conflicts between two competing objectives: maintaining existing knowledge (preservation) and incorporating new information (editing). During gradient-based optimization, these conflicting objectives can lead to imbalanced update directions, where one gradient dominates, ultimately resulting in suboptimal learning dynamics. To address this challenge, we propose a balanced knowledge editing framework inspired by Nash bargaining theory. Our method guides the optimization process toward a Pareto stationary point, ensuring an equilibrium solution wherein any deviation from the final state would degrade the overall performance with respect to both objectives. This guarantees optimality in preserving prior knowledge while integrating new information. We empirically validate the effectiveness of our approach across a range of evaluation metrics on standard benchmark datasets. Extensive experiments show that our method consistently outperforms state-of-the-art techniques, achieving a superior balance between knowledge preservation and update accuracy.

Xi Wang, Quan Shi, Zenghui Ding, Jianqing Gao, Xianjun Yang

The illusion phenomenon of large language models (LLMs) is the core obstacle to their reliable deployment. This article formalizes the large language model as a probabilistic Turing machine by constructing a "computational necessity hierarchy", and for the first time proves the illusions are inevitable on diagonalization, incomputability, and information theory boundaries supported by the new "learner pump lemma". However, we propose two "escape routes": one is to model Retrieval Enhanced Generations (RAGs) as oracle machines, proving their absolute escape through "computational jumps", providing the first formal theory for the effectiveness of RAGs; The second is to formalize continuous learning as an "internalized oracle" mechanism and implement this path through a novel neural game theory framework.Finally, this article proposes a feasible new principle for artificial intelligence security - Computational Class Alignment (CCA), which requires strict matching between task complexity and the actual computing power of the system, providing theoretical support for the secure application of artificial intelligence.