arXiv subjects

Enoch Hyunwook Kang

Publications and source records attributed to Enoch Hyunwook Kang.

7 recordsLinked to original sources

arXiv2026

Demystifying the Unreasonable Effectiveness of Greedy Alignment Methods

Greedy AI alignment methods based on Bradley-Terry reward-model fitting and KL-regularized Alignment are remarkably effective in both offline and online preference-learning pipelines, yet existing O(log T) online and O(1/epsilon) offline KL-regularized regret rate guarantees can seem pessimistic relative to their empirical performance. We argue that this mismatch reflects a distinction between two learning targets: KL-regularized regret measures how fast we can recover the KL-regularized aligned policy, whereas we often only ask how fast we can learn the reward function that induces the correct best response. To isolate this reward-learning component, we study the traditional temperature-zero regret criterion, which evaluates only the top-ranked response at inference time. Under realizability, compactness, and a uniform best-response margin over the fitted reward class, we prove that exact reference-logged offline RLHF achieves exponentially small expected temperature-zero regret, and therefore reaches expected regret at most epsilon with O(log(1ε)) offline comparisons. We also prove that exact greedy online RLHF achieves bounded O(1) cumulative temperature-zero regret; the same guarantee extends to its equivalent pairwise online DPO formulation. By separating selector identification from full soft-policy recovery, our results provide a sharper theoretical explanation for the empirical efficiency of greedy alignment.

cs.LG↗

arXiv2026

A Lecture Note on Offline RL and IRL, Part II: Foundations of Inverse Reinforcement Learning and Dynamic Discrete Choice Models

In the forward reinforcement-learning problem, the reward is fixed and known; the learner is asked to find a good policy or value function. Here we turn the question around. Given offline data generated by an expert, can we recover the reward the expert was optimizing? This is the inverse reinforcement learning problem, and remarkably, two communities, structural econometricians studying dynamic discrete choice (DDC) and machine learners studying entropy-regularized IRL, have been working on exactly the same probabilistic model under different names. We begin by proving their equivalence. We then develop the classical identification result of Magnac and Thesmar and the classical computational paradigms that grew out of it: Rust's nested fixed-point algorithm, the conditional-choice-probability approach of Hotz and Miller, and the two temporal-difference approaches of Adusumilli and Eckardt: linear semi-gradient TD and approximate value iteration. Each route has its limits: dimensionality, transition-kernel estimation, the deadly triad, or projected fixed-point bias. We then walk through the modern ML/IRL strand: adversarial IRL, occupancy matching, IQ-Learn, and offline ML-IRL, deriving each method's actual objective and stating precisely what it does and does not identify. We close with the empirical-risk-minimization framework of Kang et al., which yields a gradient-based estimator for offline IRL/DDC.

cs.LG↗

arXiv2026

Personalized Alignment Revisited: The Necessity and Sufficiency of User Diversity

Personalized alignment aims to adapt large language models to heterogeneous user preferences, yet the precise theoretical conditions for its statistical efficiency have not been formally established. This paper characterizes the conditions under which personalized alignment achieves O(1) online regret and log(1/epsilon) offline sample complexity. We show that these optimal rates depend on a specific user-diversity condition: the population of user-specific heads must span the latent reward directions that can alter the optimal response. We prove that this condition is both necessary and sufficient. When it holds, simple greedy algorithms achieve benchmark efficiency; when it fails, every learner in a natural admissible class incurs at least logarithmic regret. Our results identify user diversity as the fundamental driver of personalized identifiability.

cs.LG↗

arXiv2026

Reasonably reasoning AI agents can avoid game-theoretic failures in zero-shot, provably

As autonomous AI agents increasingly mediate online platform markets, a fundamental question emerges: do these markets generate stable strategic outcomes? In repeated strategic environments, the Nash equilibrium provides a natural benchmark for this stability. However, empirical evidence on off-the-shelf LLM agents is mixed, leaving it unclear whether independently deployed agents can converge to equilibrium behavior without explicit strategic post-training. In this paper, we provide an affirmative answer. Extending the Bayesian learning literature in theoretical economics, we prove that AI agents, acting as Bayesian posterior samplers rather than expected utility maximizers, are guaranteed to eventually become weakly close to a Nash equilibrium in infinitely repeated games. We further extend this analysis to settings in which stage payoffs are unknown ex ante, and agents observe only their privately realized stochastic payoffs, and obtain the same convergence guarantees. Finally, we empirically evaluate these theoretical implications across five repeated-game environments, ranging from the Prisoner's Dilemma to marketing promotion games. Taken together, our findings suggest that strategic stability in AI-mediated markets can emerge from the intrinsic reasoning and learning properties of modern AI agents, without the need for unrealistic universal fine-tuning.

cs.AI↗

arXiv2026

LLM Personas as a Substitute for Field Experiments in Method Benchmarking

Field experiments (A/B tests) are often the most credible benchmark for methods (algorithms) in societal systems, but their cost and latency bottleneck rapid methodological progress. LLM-based persona simulation offers a cheap synthetic alternative, yet it is unclear whether replacing humans with personas preserves the benchmark interface that adaptive methods optimize against. We prove an if-and-only-if characterization: when (i) methods observe only the aggregate outcome (aggregate-only observation) and (ii) evaluation depends only on the submitted artifact and not on the method's identity or provenance (method-blind evaluation), swapping humans for personas is just panel change from the method's point of view, indistinguishable from changing the evaluation population (e.g., New York to Jakarta). Furthermore, we move from validity to usefulness: we define an information-theoretic discriminability of the induced aggregate channel and show that making persona benchmarking as decision-relevant as a field experiment is fundamentally a sample-size question, yielding explicit bounds on the number of independent persona evaluations required to reliably distinguish meaningfully different methods at a chosen resolution.

cs.AI↗

arXiv2026

TextBO: Bayesian Optimization in Language Space for Eval-Efficient Self-Improving AI

Large Language Models (LLMs) have enabled self-improving AI systems that iteratively generate, evaluate, and refine their outcomes. Recent studies show that prompt-optimization-based self-improvement can outperform state-of-the-art reinforcement-learning fine-tuning of LLMs, but performance is typically measured by generation efficiency. However, in many applications, the constraint is evaluation efficiency: obtaining reliable feedback is far more costly than generating candidates. To optimize for evaluation efficiency, we extend Upper Confidence Bound-Bayesian Optimization (UCB-BO), a framework known for optimal evaluation-efficiency guarantees, to the language domain. Doing so is challenging for two reasons: (i) gradients needed for UCB-BO are ill-defined in discrete prompt space; and (ii) UCB-style exploration relies on a surrogate model and acquisition function, which only live implicitly in the LLM. We overcome these challenges by proving that combining simple textual gradients (LLM-proposed local edits) with the Best-of-N selection strategy statistically emulates ascent along the gradient of the canonical UCB acquisition function. Based on this result, we propose TextBO, a simple, evaluation-efficient self-improving algorithm that operates purely in language space without explicit surrogates or calibrated uncertainty models. We empirically validate TextBO on automated ad-alignment tasks using a persona-induced preference distribution, demonstrating superior performance per evaluation compared to strong baselines such as Best-of-N and GEPA. We also evaluate TextBO's Best-of-N multi-step textual-gradient mechanism on agentic AI benchmarks by augmenting GEPA with it and show that it significantly outperforms standard GEPA. In sum, TextBO is a simple and principled framework for AI self-improving system design that bridges prompt optimization with classical Bayesian optimization.

cs.AI↗

arXiv2023

Bounded (O(1)) Regret Recommendation Learning via Synthetic Controls Oracle

In online exploration systems where users with fixed preferences repeatedly arrive, it has recently been shown that O(1), i.e., bounded regret, can be achieved when the system is modeled as a linear contextual bandit. This result may be of interest for recommender systems, where the popularity of their items is often short-lived, as the exploration itself may be completed quickly before potential long-run non-stationarities come into play. However, in practice, exact knowledge of the linear model is difficult to justify. Furthermore, potential existence of unobservable covariates, uneven user arrival rates, interpretation of the necessary rank condition, and users opting out of private data tracking all need to be addressed for practical recommender system applications. In this work, we conduct a theoretical study to address all these issues while still achieving bounded regret. Aside from proof techniques, the key differentiating assumption we make here is the presence of effective Synthetic Control Methods (SCM), which are shown to be a practical relaxation of the exact linear model knowledge assumption. We verify our theoretical bounded regret result using a minimal simulation experiment.

cs.LG↗