Beyond Reverse KL: Generalizing Direct Preference Optimization with Diverse Divergence Constraints

The increasing capabilities of large language models (LLMs) raise opportunities for artificial general intelligence but concurrently amplify safety concerns, such as potential misuse of AI systems, necessitating effective AI alignment. Reinforcement Learning from Human Feedback (RLHF) has emerged as a promising pathway towards AI alignment but brings forth challenges due to its complexity and dependence on a separate reward model. Direct Preference Optimization (DPO) has been proposed as an alternative, and it remains equivalent to RLHF under the reverse KL regularization constraint. This paper presents $f$-DPO, a generalized approach to DPO by incorporating diverse divergence constraints. We show that under certain $f$-divergences, including Jensen-Shannon divergence, forward KL divergences and $\alpha$-divergences, the complex relationship between the reward and optimal policy can also be simplified by addressing the Karush-Kuhn-Tucker conditions. This eliminates the need for estimating the normalizing constant in the Bradley-Terry model and enables a tractable mapping between the reward function and the optimal policy. Our approach optimizes LLMs to align with human preferences in a more efficient and supervised manner under a broad set of divergence constraints. Empirically, adopting these divergences ensures a balance between alignment performance and generation diversity. Importantly, $f$-DPO outperforms PPO-based methods in divergence efficiency, and divergence constraints directly influence expected calibration error (ECE).

BitCoin: Bidirectional Tagging and Supervised Contrastive Learning based Joint Relational Triple Extraction Framework

Relation triple extraction (RTE) is an essential task in information extraction and knowledge graph construction. Despite recent advancements, existing methods still exhibit certain limitations. They just employ generalized pre-trained models and do not consider the specificity of RTE tasks. Moreover, existing tagging-based approaches typically decompose the RTE task into two subtasks, initially identifying subjects and subsequently identifying objects and relations. They solely focus on extracting relational triples from subject to object, neglecting that once the extraction of a subject fails, it fails in extracting all triples associated with that subject. To address these issues, we propose BitCoin, an innovative Bidirectional tagging and supervised Contrastive learning based joint relational triple extraction framework. Specifically, we design a supervised contrastive learning method that considers multiple positives per anchor rather than restricting it to just one positive. Furthermore, a penalty term is introduced to prevent excessive similarity between the subject and object. Our framework implements taggers in two directions, enabling triples extraction from subject to object and object to subject. Experimental results show that BitCoin achieves state-of-the-art results on the benchmark datasets and significantly improves the F1 score on Normal, SEO, EPO, and multiple relation extraction tasks.

Settling the Sample Complexity of Online Reinforcement Learning

A central issue lying at the heart of online reinforcement learning (RL) is data efficiency. While a number of recent works achieved asymptotically minimal regret in online RL, the optimality of these results is only guaranteed in a ``large-sample'' regime, imposing enormous burn-in cost in order for their algorithms to operate optimally. How to achieve minimax-optimal regret without incurring any burn-in cost has been an open problem in RL theory. We settle this problem for the context of finite-horizon inhomogeneous Markov decision processes. Specifically, we prove that a modified version of Monotonic Value Propagation (MVP), a model-based algorithm proposed by \cite{zhang2020reinforcement}, achieves a regret on the order of (modulo log factors) \begin{equation*} \min\big\{ \sqrt{SAH^3K}, \,HK \big\}, \end{equation*} where $S$ is the number of states, $A$ is the number of actions, $H$ is the planning horizon, and $K$ is the total number of episodes. This regret matches the minimax lower bound for the entire range of sample size $K\geq 1$, essentially eliminating any burn-in requirement. It also translates to a PAC sample complexity (i.e., the number of episodes needed to yield $\varepsilon$-accuracy) of $\frac{SAH^3}{\varepsilon^2}$ up to log factor, which is minimax-optimal for the full $\varepsilon$-range. Further, we extend our theory to unveil the influences of problem-dependent quantities like the optimal value/cost and certain variances. The key technical innovation lies in the development of a new regret decomposition strategy and a novel analysis paradigm to decouple complicated statistical dependency -- a long-standing challenge facing the analysis of online RL in the sample-hungry regime.

Learning Regions of Interest for Bayesian Optimization with Adaptive Level-Set Estimation

We study Bayesian optimization (BO) in high-dimensional and non-stationary scenarios. Existing algorithms for such scenarios typically require extensive hyperparameter tuning, which limits their practical effectiveness. We propose a framework, called BALLET, which adaptively filters for a high-confidence region of interest (ROI) as a superlevel-set of a nonparametric probabilistic model such as a Gaussian process (GP). Our approach is easy to tune, and is able to focus on local region of the optimization space that can be tackled by existing BO methods. The key idea is to use two probabilistic models: a coarse GP to identify the ROI, and a localized GP for optimization within the ROI. We show theoretically that BALLET can efficiently shrink the search space, and can exhibit a tighter regret bound than standard BO without ROI filtering. We demonstrate empirically the effectiveness of BALLET on both synthetic and real-world optimization tasks.

Active Policy Improvement from Multiple Black-box Oracles

Reinforcement learning (RL) has made significant strides in various complex domains. However, identifying an effective policy via RL often necessitates extensive exploration. Imitation learning aims to mitigate this issue by using expert demonstrations to guide exploration. In real-world scenarios, one often has access to multiple suboptimal black-box experts, rather than a single optimal oracle. These experts do not universally outperform each other across all states, presenting a challenge in actively deciding which oracle to use and in which state. We introduce MAPS and MAPS-SE, a class of policy improvement algorithms that perform imitation learning from multiple suboptimal oracles. In particular, MAPS actively selects which of the oracles to imitate and improve their value function estimates, and MAPS-SE additionally leverages an active state exploration criterion to determine which states one should explore. We provide a comprehensive theoretical analysis and demonstrate that MAPS and MAPS-SE enjoy sample efficiency advantage over the state-of-the-art policy improvement algorithms. Empirical results show that MAPS-SE significantly accelerates policy optimization via state-wise imitation learning from multiple oracles across a broad spectrum of control tasks in the DeepMind Control Suite. Our code is publicly available at: https://github.com/ripl/maps.

Towards Faster Non-Asymptotic Convergence for Diffusion-Based Generative Models

Diffusion models, which convert noise into new data instances by learning to reverse a Markov diffusion process, have become a cornerstone in contemporary generative modeling. While their practical power has now been widely recognized, the theoretical underpinnings remain far from mature. In this work, we develop a suite of non-asymptotic theory towards understanding the data generation process of diffusion models in discrete time, assuming access to reliable estimates of the (Stein) score functions. For a popular deterministic sampler (based on the probability flow ODE), we establish a convergence rate proportional to $1/T$ (with $T$ the total number of steps), improving upon past results; for another mainstream stochastic sampler (i.e., a type of the denoising diffusion probabilistic model (DDPM)), we derive a convergence rate proportional to $1/\sqrt{T}$, matching the state-of-the-art theory. Our theory imposes only minimal assumptions on the target data distribution (e.g., no smoothness assumption is imposed), and is developed based on an elementary yet versatile non-asymptotic approach without resorting to toolboxes for SDEs and ODEs. Further, we design two accelerated variants, improving the convergence to $1/T^2$ for the ODE-based sampler and $1/T$ for the DDPM-type sampler, which might be of independent theoretical and empirical interest.

The Curious Price of Distributional Robustness in Reinforcement Learning with a Generative Model

This paper investigates model robustness in reinforcement learning (RL) to reduce the sim-to-real gap in practice. We adopt the framework of distributionally robust Markov decision processes (RMDPs), aimed at learning a policy that optimizes the worst-case performance when the deployed environment falls within a prescribed uncertainty set around the nominal MDP. Despite recent efforts, the sample complexity of RMDPs remained mostly unsettled regardless of the uncertainty set in use. It was unclear if distributional robustness bears any statistical consequences when benchmarked against standard RL. Assuming access to a generative model that draws samples based on the nominal MDP, we characterize the sample complexity of RMDPs when the uncertainty set is specified via either the total variation (TV) distance or $\chi^2$ divergence. The algorithm studied here is a model-based method called {\em distributionally robust value iteration}, which is shown to be near-optimal for the full range of uncertainty levels. Somewhat surprisingly, our results uncover that RMDPs are not necessarily easier or harder to learn than standard MDPs. The statistical consequence incurred by the robustness requirement depends heavily on the size and shape of the uncertainty set: in the case w.r.t.~the TV distance, the minimax sample complexity of RMDPs is always smaller than that of standard MDPs; in the case w.r.t.~the $\chi^2$ divergence, the sample complexity of RMDPs can often far exceed the standard MDP counterpart.

Reward-agnostic Fine-tuning: Provable Statistical Benefits of Hybrid Reinforcement Learning

This paper studies tabular reinforcement learning (RL) in the hybrid setting, which assumes access to both an offline dataset and online interactions with the unknown environment. A central question boils down to how to efficiently utilize online data collection to strengthen and complement the offline dataset and enable effective policy fine-tuning. Leveraging recent advances in reward-agnostic exploration and model-based offline RL, we design a three-stage hybrid RL algorithm that beats the best of both worlds -- pure offline RL and pure online RL -- in terms of sample complexities. The proposed algorithm does not require any reward information during data collection. Our theory is developed based on a new notion called single-policy partial concentrability, which captures the trade-off between distribution mismatch and miscoverage and guides the interplay between offline and online data.

Efficient Online Decision Tree Learning with Active Feature Acquisition

Constructing decision trees online is a classical machine learning problem. Existing works often assume that features are readily available for each incoming data point. However, in many real world applications, both feature values and the labels are unknown a priori and can only be obtained at a cost. For example, in medical diagnosis, doctors have to choose which tests to perform (i.e., making costly feature queries) on a patient in order to make a diagnosis decision (i.e., predicting labels). We provide a fresh perspective to tackle this practical challenge. Our framework consists of an active planning oracle embedded in an online learning scheme for which we investigate several information acquisition functions. Specifically, we employ a surrogate information acquisition function based on adaptive submodularity to actively query feature values with a minimal cost, while using a posterior sampling scheme to maintain a low regret for online prediction. We demonstrate the efficiency and effectiveness of our framework via extensive experiments on various real-world datasets. Our framework also naturally adapts to the challenging setting of online learning with concept drift and is shown to be competitive with baseline models while being more flexible.

Minimax-Optimal Reward-Agnostic Exploration in Reinforcement Learning

This paper studies reward-agnostic exploration in reinforcement learning (RL) -- a scenario where the learner is unware of the reward functions during the exploration stage -- and designs an algorithm that improves over the state of the art. More precisely, consider a finite-horizon non-stationary Markov decision process with $S$ states, $A$ actions, and horizon length $H$, and suppose that there are no more than a polynomial number of given reward functions of interest. By collecting an order of \begin{align*} \frac{SAH^3}{\varepsilon^2} \text{ sample episodes (up to log factor)} \end{align*} without guidance of the reward information, our algorithm is able to find $\varepsilon$-optimal policies for all these reward functions, provided that $\varepsilon$ is sufficiently small. This forms the first reward-agnostic exploration scheme in this context that achieves provable minimax optimality. Furthermore, once the sample size exceeds $\frac{S^2AH^3}{\varepsilon^2}$ episodes (up to log factor), our algorithm is able to yield $\varepsilon$ accuracy for arbitrarily many reward functions (even when they are adversarially designed), a task commonly dubbed as ``reward-free exploration.'' The novelty of our algorithm design draws on insights from offline RL: the exploration scheme attempts to maximize a critical reward-agnostic quantity that dictates the performance of offline RL, while the policy learning paradigm leverages ideas from sample-optimal offline RL paradigms.