Large language models (LLMs) have demonstrated remarkable in-context learning (ICL) abilities. However, existing theoretical analysis of ICL primarily exhibits two limitations: (a) Limited i.i.d. Setting. Most studies focus on supervised function learning tasks where prompts are constructed with i.i.d. input-label pairs. This i.i.d. assumption diverges significantly from real language learning scenarios where prompt tokens are interdependent. (b) Lack of Emergence Explanation. Most literature answers what ICL does from an implicit optimization perspective but falls short in elucidating how ICL emerges and the impact of pre-training phase on ICL. In our paper, to extend (a), we adopt a more practical paradigm, auto-regressive next-token prediction (AR-NTP), which closely aligns with the actual training of language models. Specifically, within AR-NTP, we emphasize prompt token-dependency, which involves predicting each subsequent token based on the preceding sequence. To address (b), we formalize a systematic pre-training and ICL framework, highlighting the layer-wise structure of sequences and topics, alongside a two-level expectation. In conclusion, we present data-dependent, topic-dependent and optimization-dependent PAC-Bayesian generalization bounds for pre-trained LLMs, investigating that ICL emerges from the generalization of sequences and topics. Our theory is supported by experiments on numerical linear dynamic systems, synthetic GINC and real-world language datasets.

Motivated by the great success of Bayesian additive regression trees (BART) on regression, we propose a nonparametric Bayesian approach for the function-on-scalar regression problem, termed as Functional BART (FBART). Utilizing spline-based function representation and tree-based domain partition model, FBART offers great flexibility in characterizing the complex and heterogeneous relationship between the response curve and scalar covariates. We devise a tailored Bayesian backfitting algorithm for estimating the parameters in the FBART model. Furthermore, we introduce an FBART model with shape constraints on the response curve, enhancing estimation and prediction performance when prior shape information of response curves is available. By incorporating a shape-constrained prior, we ensure that the posterior samples of the response curve satisfy the required shape constraints (e.g., monotonicity and/or convexity). Our proposed FBART model and its shape-constrained version are the new advances of BART models for functional data. Under certain regularity conditions, we derive the posterior convergence results for both FBART and its shape-constrained version. Finally, the superiority of the proposed methods over other competitive counterparts is validated through simulation experiments under various settings and analyses of two real datasets.

We develop an analysis of Thompson sampling for online learning under full feedback - also known as prediction with expert advice - where the learner's prior is defined over the space of an adversary's future actions, rather than the space of experts. We show regret decomposes into regret the learner expected a priori, plus a prior-robustness-type term we call excess regret. In the classical finite-expert setting, this recovers optimal rates. As an initial step towards practical online learning in settings with a potentially-uncountably-infinite number of experts, we show that Thompson sampling with a certain Gaussian process prior widely-used in the Bayesian optimization literature has a $\mathcal{O}(\beta\sqrt{T\log(1+\lambda)})$ rate against a $\beta$-bounded $\lambda$-Lipschitz adversary.

Emergency Response Time (ERT) is crucial for urban safety, measuring cities' ability to handle medical, fire, and crime emergencies. In NYC, medical ERT increased 72% from 7.89 minutes in 2014 to 14.27 minutes in 2024, with half of delays due to Emergency Vehicle (EMV) travel times. Each minute's delay in stroke response costs 2 million brain cells, while cardiac arrest survival drops 7-10% per minute. This dissertation advances EMV facilitation through three contributions. First, EMVLight, a decentralized multi-agent reinforcement learning framework, integrates EMV routing with traffic signal pre-emption. It achieved 42.6% faster EMV travel times and 23.5% improvement for other vehicles. Second, the Dynamic Queue-Jump Lane system uses Multi-Agent Proximal Policy Optimization for coordinated lane-clearing in mixed autonomous and human-driven traffic, reducing EMV travel times by 40%. Third, an equity study of NYC Emergency Medical Services revealed disparities across boroughs: Staten Island faces delays due to sparse signalized intersections, while Manhattan struggles with congestion. Solutions include optimized EMS stations and improved intersection designs. These contributions enhance EMV mobility and emergency service equity, offering insights for policymakers and urban planners to develop safer, more efficient transportation systems.

A tenet of reinforcement learning is that rewards are always observed by the agent. However, this is not true in many realistic settings, e.g., a human observer may not always be able to provide rewards, a sensor to observe rewards may be limited or broken, or rewards may be unavailable during deployment. Monitored Markov decision processes (Mon-MDPs) have recently been proposed as a model of such settings. Yet, Mon-MDP algorithms developed thus far do not fully exploit the problem structure, cannot take advantage of a known monitor, have no worst-case guarantees for ``unsolvable'' Mon-MDPs without specific initialization, and only have asymptotic proofs of convergence. This paper makes three contributions. First, we introduce a model-based algorithm for Mon-MDPs that addresses all of these shortcomings. The algorithm uses two instances of model-based interval estimation, one to guarantee that observable rewards are indeed observed, and another to learn the optimal policy. Second, empirical results demonstrate these advantages, showing faster convergence than prior algorithms in over two dozen benchmark settings, and even more dramatic improvements when the monitor process is known. Third, we present the first finite-sample bound on performance and show convergence to an optimal worst-case policy when some rewards are never observable.

We study a ranking and selection (R&S) problem when all solutions share common parametric Bayesian input models updated with the data collected from multiple independent data-generating sources. Our objective is to identify the best system by designing a sequential sampling algorithm that collects input and simulation data given a budget. We adopt the most probable best (MPB) as the estimator of the optimum and show that its posterior probability of optimality converges to one at an exponential rate as the sampling budget increases. Assuming that the input parameters belong to a finite set, we characterize the $\epsilon$-optimal static sampling ratios for input and simulation data that maximize the convergence rate. Using these ratios as guidance, we propose the optimal sampling algorithm for R&S (OSAR) that achieves the $\epsilon$-optimal ratios almost surely in the limit. We further extend OSAR by adopting the kernel ridge regression to improve the simulation output mean prediction. This not only improves OSAR's finite-sample performance, but also lets us tackle the case where the input parameters lie in a continuous space with a strong consistency guarantee for finding the optimum. We numerically demonstrate that OSAR outperforms a state-of-the-art competitor.

State-of-the-art data stream mining in supervised classification has traditionally relied on ensembles of incremental decision trees. However, the emergence of large tabular models, i.e., transformers designed for structured numerical data, marks a significant paradigm shift. These models move beyond traditional weight updates, instead employing in-context learning through prompt tuning. By using on-the-fly sketches to summarize unbounded streaming data, one can feed this information into a pre-trained model for efficient processing. This work bridges advancements from both areas, highlighting how transformers' implicit meta-learning abilities, pre-training on drifting natural data, and reliance on context optimization directly address the core challenges of adaptive learning in dynamic environments. Exploring real-time model adaptation, this research demonstrates that TabPFN, coupled with a simple sliding memory strategy, consistently outperforms ensembles of Hoeffding trees across all non-stationary benchmarks. Several promising research directions are outlined in the paper. The authors urge the community to explore these ideas, offering valuable opportunities to advance in-context stream learning.

Training large language models to predict beyond their training context lengths has drawn much attention in recent years, yet the principles driving such behavior of length generalization remain underexplored. We propose a new theoretical framework to study length generalization for the next-token prediction task, as performed by decoder-only transformers. Conceptually, we show that length generalization occurs as long as each predicted token depends on a small (fixed) number of previous tokens. We formalize such tasks via a notion we call $k$-sparse planted correlation distributions, and show that an idealized model of transformers which generalize attention heads successfully length-generalize on such tasks. As a bonus, our theoretical model justifies certain techniques to modify positional embeddings which have been introduced to improve length generalization, such as position coupling. We support our theoretical results with experiments on synthetic tasks and natural language, which confirm that a key factor driving length generalization is a ``sparse'' dependency structure of each token on the previous ones. Inspired by our theory, we introduce Predictive Position Coupling, which trains the transformer to predict the position IDs used in a positional coupling approach. Predictive Position Coupling thereby allows us to broaden the array of tasks to which position coupling can successfully be applied to achieve length generalization.

Ensuring the robustness of deep reinforcement learning (DRL) agents against adversarial attacks is critical for their trustworthy deployment. Recent research highlights the challenges of achieving state-adversarial robustness and suggests that an optimal robust policy (ORP) does not always exist, complicating the enforcement of strict robustness constraints. In this paper, we further explore the concept of ORP. We first introduce the Intrinsic State-adversarial Markov Decision Process (ISA-MDP), a novel formulation where adversaries cannot fundamentally alter the intrinsic nature of state observations. ISA-MDP, supported by empirical and theoretical evidence, universally characterizes decision-making under state-adversarial paradigms. We rigorously prove that within ISA-MDP, a deterministic and stationary ORP exists, aligning with the Bellman optimal policy. Our findings theoretically reveal that improving DRL robustness does not necessarily compromise performance in natural environments. Furthermore, we demonstrate the necessity of infinity measurement error (IME) in both $Q$-function and probability spaces to achieve ORP, unveiling vulnerabilities of previous DRL algorithms that rely on $1$-measurement errors. Motivated by these insights, we develop the Consistent Adversarial Robust Reinforcement Learning (CAR-RL) framework, which optimizes surrogates of IME. We apply CAR-RL to both value-based and policy-based DRL algorithms, achieving superior performance and validating our theoretical analysis.

Solving complex long-horizon robotic manipulation problems requires sophisticated high-level planning capabilities, the ability to reason about the physical world, and reactively choose appropriate motor skills. Vision-language models (VLMs) pretrained on Internet data could in principle offer a framework for tackling such problems. However, in their current form, VLMs lack both the nuanced understanding of intricate physics required for robotic manipulation and the ability to reason over long horizons to address error compounding issues. In this paper, we introduce a novel test-time computation framework that enhances VLMs' physical reasoning capabilities for multi-stage manipulation tasks. At its core, our approach iteratively improves a pretrained VLM with a "reflection" mechanism - it uses a generative model to imagine future world states, leverages these predictions to guide action selection, and critically reflects on potential suboptimalities to refine its reasoning. Experimental results demonstrate that our method significantly outperforms several state-of-the-art commercial VLMs as well as other post-training approaches such as Monte Carlo Tree Search (MCTS). Videos are available at https://reflect-vlm.github.io.