Exploration-Driven Policy Optimization in RLHF: Theoretical Insights on Efficient Data Utilization

Reinforcement Learning from Human Feedback (RLHF) has achieved impressive empirical successes while relying on a small amount of human feedback. However, there is limited theoretical justification for this phenomenon. Additionally, most recent studies focus on value-based algorithms despite the recent empirical successes of policy-based algorithms. In this work, we consider an RLHF algorithm based on policy optimization (PO-RLHF). The algorithm is based on the popular Policy Cover-Policy Gradient (PC-PG) algorithm, which assumes knowledge of the reward function. In PO-RLHF, knowledge of the reward function is not assumed and the algorithm relies on trajectory-based comparison feedback to infer the reward function. We provide performance bounds for PO-RLHF with low query complexity, which provides insight into why a small amount of human feedback may be sufficient to get good performance with RLHF. A key novelty is our trajectory-level elliptical potential analysis technique used to infer reward function parameters when comparison queries rather than reward observations are used. We provide and analyze algorithms in two settings: linear and neural function approximation, PG-RLHF and NN-PG-RLHF, respectively.

Cascading Reinforcement Learning

Cascading bandits have gained popularity in recent years due to their applicability to recommendation systems and online advertising. In the cascading bandit model, at each timestep, an agent recommends an ordered subset of items (called an item list) from a pool of items, each associated with an unknown attraction probability. Then, the user examines the list, and clicks the first attractive item (if any), and after that, the agent receives a reward. The goal of the agent is to maximize the expected cumulative reward. However, the prior literature on cascading bandits ignores the influences of user states (e.g., historical behaviors) on recommendations and the change of states as the session proceeds. Motivated by this fact, we propose a generalized cascading RL framework, which considers the impact of user states and state transition into decisions. In cascading RL, we need to select items not only with large attraction probabilities but also leading to good successor states. This imposes a huge computational challenge due to the combinatorial action space. To tackle this challenge, we delve into the properties of value functions, and design an oracle BestPerm to efficiently find the optimal item list. Equipped with BestPerm, we develop two algorithms CascadingVI and CascadingBPI, which are both computationally-efficient and sample-efficient, and provide near-optimal regret and sample complexity guarantees. Furthermore, we present experiments to show the improved computational and sample efficiencies of our algorithms compared to straightforward adaptations of existing RL algorithms in practice.

Provably Efficient Iterated CVaR Reinforcement Learning with Function Approximation

Risk-sensitive reinforcement learning (RL) aims to optimize policies that balance the expected reward and risk. In this paper, we investigate a novel risk-sensitive RL formulation with an Iterated Conditional Value-at-Risk (CVaR) objective under linear and general function approximations. This new formulation, named ICVaR-RL with function approximation, provides a principled way to guarantee safety at each decision step. For ICVaR-RL with linear function approximation, we propose a computationally efficient algorithm ICVaR-L, which achieves an $\widetilde{O}(\sqrt{\alpha^{-(H+1)}(d^2H^4+dH^6)K})$ regret, where $\alpha$ is the risk level, $d$ is the dimension of state-action features, $H$ is the length of each episode, and $K$ is the number of episodes. We also establish a matching lower bound $\Omega(\sqrt{\alpha^{-(H-1)}d^2K})$ to validate the optimality of ICVaR-L with respect to $d$ and $K$. For ICVaR-RL with general function approximation, we propose algorithm ICVaR-G, which achieves an $\widetilde{O}(\sqrt{\alpha^{-(H+1)}DH^4K})$ regret, where $D$ is a dimensional parameter that depends on the eluder dimension and covering number. Furthermore, our analysis provides several novel techniques for risk-sensitive RL, including an efficient approximation of the CVaR operator, a new ridge regression with CVaR-adapted features, and a refined elliptical potential lemma.

Multi-task Representation Learning for Pure Exploration in Linear Bandits

Despite the recent success of representation learning in sequential decision making, the study of the pure exploration scenario (i.e., identify the best option and minimize the sample complexity) is still limited. In this paper, we study multi-task representation learning for best arm identification in linear bandits (RepBAI-LB) and best policy identification in contextual linear bandits (RepBPI-CLB), two popular pure exploration settings with wide applications, e.g., clinical trials and web content optimization. In these two problems, all tasks share a common low-dimensional linear representation, and our goal is to leverage this feature to accelerate the best arm (policy) identification process for all tasks. For these problems, we design computationally and sample efficient algorithms DouExpDes and C-DouExpDes, which perform double experimental designs to plan optimal sample allocations for learning the global representation. We show that by learning the common representation among tasks, our sample complexity is significantly better than that of the native approach which solves tasks independently. To the best of our knowledge, this is the first work to demonstrate the benefits of representation learning for multi-task pure exploration.

Dueling Bandits: From Two-dueling to Multi-dueling

We study a general multi-dueling bandit problem, where an agent compares multiple options simultaneously and aims to minimize the regret due to selecting suboptimal arms. This setting generalizes the traditional two-dueling bandit problem and finds many real-world applications involving subjective feedback on multiple options. We start with the two-dueling bandit setting and propose two efficient algorithms, DoublerBAI and MultiSBM-Feedback. DoublerBAI provides a generic schema for translating known results on best arm identification algorithms to the dueling bandit problem, and achieves a regret bound of $O(\ln T)$. MultiSBM-Feedback not only has an optimal $O(\ln T)$ regret, but also reduces the constant factor by almost a half compared to benchmark results. Then, we consider the general multi-dueling case and develop an efficient algorithm MultiRUCB. Using a novel finite-time regret analysis for the general multi-dueling bandit problem, we show that MultiRUCB also achieves an $O(\ln T)$ regret bound and the bound tightens as the capacity of the comparison set increases. Based on both synthetic and real-world datasets, we empirically demonstrate that our algorithms outperform existing algorithms.

Risk-Sensitive Reinforcement Learning: Iterated CVaR and the Worst Path

In this paper, we study a novel episodic risk-sensitive Reinforcement Learning (RL) problem, named Iterated CVaR RL, where the objective is to maximize the tail of the reward-to-go at each step. Different from existing risk-aware RL formulations, Iterated CVaR RL focuses on safety-at-all-time, by enabling the agent to tightly control the risk of getting into catastrophic situations at each stage, and is applicable to important risk-sensitive tasks that demand strong safety guarantees throughout the decision process, such as autonomous driving, clinical treatment planning and robotics. We investigate Iterated CVaR RL with two performance metrics, i.e., Regret Minimization and Best Policy Identification. For both metrics, we design efficient algorithms ICVaR-RM and ICVaR-BPI, respectively, and provide matching upper and lower bounds with respect to the number of episodes $K$. We also investigate an interesting limiting case of Iterated CVaR RL, called Worst Path RL, where the objective becomes to maximize the minimum possible cumulative reward, and propose an efficient algorithm with constant upper and lower bounds. Finally, the techniques we develop for bounding the change of CVaR due to the value function shift and decomposing the regret via a distorted visitation distribution are novel and can find applications in other risk-sensitive online learning problems.

Branching Reinforcement Learning

In this paper, we propose a novel Branching Reinforcement Learning (Branching RL) model, and investigate both Regret Minimization (RM) and Reward-Free Exploration (RFE) metrics for this model. Unlike standard RL where the trajectory of each episode is a single $H$-step path, branching RL allows an agent to take multiple base actions in a state such that transitions branch out to multiple successor states correspondingly, and thus it generates a tree-structured trajectory. This model finds important applications in hierarchical recommendation systems and online advertising. For branching RL, we establish new Bellman equations and key lemmas, i.e., branching value difference lemma and branching law of total variance, and also bound the total variance by only $O(H^2)$ under an exponentially-large trajectory. For RM and RFE metrics, we propose computationally efficient algorithms BranchVI and BranchRFE, respectively, and derive nearly matching upper and lower bounds. Our results are only polynomial in problem parameters despite exponentially-large trajectories.

Collaborative Pure Exploration in Kernel Bandit

In this paper, we formulate a Collaborative Pure Exploration in Kernel Bandit problem (CoPE-KB), which provides a novel model for multi-agent multi-task decision making under limited communication and general reward functions, and is applicable to many online learning tasks, e.g., recommendation systems and network scheduling. We consider two settings of CoPE-KB, i.e., Fixed-Confidence (FC) and Fixed-Budget (FB), and design two optimal algorithms CoopKernelFC (for FC) and CoopKernelFB (for FB). Our algorithms are equipped with innovative and efficient kernelized estimators to simultaneously achieve computation and communication efficiency. Matching upper and lower bounds under both the statistical and communication metrics are established to demonstrate the optimality of our algorithms. The theoretical bounds successfully quantify the influences of task similarities on learning acceleration and only depend on the effective dimension of the kernelized feature space. Our analytical techniques, including data dimension decomposition, linear structured instance transformation and (communication) round-speedup induction, are novel and applicable to other bandit problems. Empirical evaluations are provided to validate our theoretical results and demonstrate the performance superiority of our algorithms.

Combinatorial Pure Exploration with Bottleneck Reward Function and its Extension to General Reward Functions

In this paper, we study the Combinatorial Pure Exploration problem with the bottleneck reward function (CPE-B) under the fixed-confidence and fixed-budget settings. In CPE-B, given a set of base arms and a collection of subsets of base arms (super arms) following certain combinatorial constraint, a learner sequentially plays (samples) a base arm and observes its random outcome, with the objective of finding the optimal super arm that maximizes its bottleneck value, defined as the minimum expected value among the base arms contained in the super arm. CPE-B captures a variety of practical scenarios such as network routing in communication networks, but it cannot be solved by the existing CPE algorithms since most of them assumed linear reward functions. For CPE-B, we present both fixed-confidence and fixed-budget algorithms, and provide the sample complexity lower bound for the fixed-confidence setting, which implies that our algorithms match the lower bound (within a logarithmic factor) for a broad family of instances. In addition, we extend CPE-B to general reward functions (CPE-G) and propose the first fixed-confidence algorithm for general non-linear reward functions with non-trivial sample complexity. Our experimental results on the top-$k$, path and matching instances demonstrate the empirical superiority of our proposed algorithms over the baselines.

Continuous Mean-Covariance Bandits

Existing risk-aware multi-armed bandit models typically focus on risk measures of individual options such as variance. As a result, they cannot be directly applied to important real-world online decision making problems with correlated options. In this paper, we propose a novel Continuous Mean-Covariance Bandit (CMCB) model to explicitly take into account option correlation. Specifically, in CMCB, there is a learner who sequentially chooses weight vectors on given options and observes random feedback according to the decisions. The agent's objective is to achieve the best trade-off between reward and risk, measured with option covariance. To capture important reward observation scenarios in practice, we consider three feedback settings, i.e., full-information, semi-bandit and full-bandit feedback. We propose novel algorithms with the optimal regrets (within logarithmic factors), and provide matching lower bounds to validate their optimalities. Our experimental results also demonstrate the superiority of the proposed algorithms. To the best of our knowledge, this is the first work that considers option correlation in risk-aware bandits and explicitly quantifies how arbitrary covariance structures impact the learning performance.