Efficient Exploration for LLMs

We present evidence of substantial benefit from efficient exploration in gathering human feedback to improve large language models. In our experiments, an agent sequentially generates queries while fitting a reward model to the feedback received. Our best-performing agent generates queries using double Thompson sampling, with uncertainty represented by an epistemic neural network. Our results demonstrate that efficient exploration enables high levels of performance with far fewer queries. Further, both uncertainty estimation and the choice of exploration scheme play critical roles.

Efficient Online Learning with Offline Datasets for Infinite Horizon MDPs: A Bayesian Approach

In this paper, we study the problem of efficient online reinforcement learning in the infinite horizon setting when there is an offline dataset to start with. We assume that the offline dataset is generated by an expert but with unknown level of competence, i.e., it is not perfect and not necessarily using the optimal policy. We show that if the learning agent models the behavioral policy (parameterized by a competence parameter) used by the expert, it can do substantially better in terms of minimizing cumulative regret, than if it doesn't do that. We establish an upper bound on regret of the exact informed PSRL algorithm that scales as $\tilde{O}(\sqrt{T})$. This requires a novel prior-dependent regret analysis of Bayesian online learning algorithms for the infinite horizon setting. We then propose an approximate Informed RLSVI algorithm that we can interpret as performing imitation learning with the offline dataset, and then performing online learning.

Sequential Best-Arm Identification with Application to Brain-Computer Interface

A brain-computer interface (BCI) is a technology that enables direct communication between the brain and an external device or computer system. It allows individuals to interact with the device using only their thoughts, and holds immense potential for a wide range of applications in medicine, rehabilitation, and human augmentation. An electroencephalogram (EEG) and event-related potential (ERP)-based speller system is a type of BCI that allows users to spell words without using a physical keyboard, but instead by recording and interpreting brain signals under different stimulus presentation paradigms. Conventional non-adaptive paradigms treat each word selection independently, leading to a lengthy learning process. To improve the sampling efficiency, we cast the problem as a sequence of best-arm identification tasks in multi-armed bandits. Leveraging pre-trained large language models (LLMs), we utilize the prior knowledge learned from previous tasks to inform and facilitate subsequent tasks. To do so in a coherent way, we propose a sequential top-two Thompson sampling (STTS) algorithm under the fixed-confidence setting and the fixed-budget setting. We study the theoretical property of the proposed algorithm, and demonstrate its substantial empirical improvement through both synthetic data analysis as well as a P300 BCI speller simulator example.

Bridging Imitation and Online Reinforcement Learning: An Optimistic Tale

In this paper, we address the following problem: Given an offline demonstration dataset from an imperfect expert, what is the best way to leverage it to bootstrap online learning performance in MDPs. We first propose an Informed Posterior Sampling-based RL (iPSRL) algorithm that uses the offline dataset, and information about the expert's behavioral policy used to generate the offline dataset. Its cumulative Bayesian regret goes down to zero exponentially fast in N, the offline dataset size if the expert is competent enough. Since this algorithm is computationally impractical, we then propose the iRLSVI algorithm that can be seen as a combination of the RLSVI algorithm for online RL, and imitation learning. Our empirical results show that the proposed iRLSVI algorithm is able to achieve significant reduction in regret as compared to two baselines: no offline data, and offline dataset but used without information about the generative policy. Our algorithm bridges online RL and imitation learning for the first time.

Leveraging Demonstrations to Improve Online Learning: Quality Matters

We investigate the extent to which offline demonstration data can improve online learning. It is natural to expect some improvement, but the question is how, and by how much? We show that the degree of improvement must depend on the quality of the demonstration data. To generate portable insights, we focus on Thompson sampling (TS) applied to a multi-armed bandit as a prototypical online learning algorithm and model. The demonstration data is generated by an expert with a given competence level, a notion we introduce. We propose an informed TS algorithm that utilizes the demonstration data in a coherent way through Bayes' rule and derive a prior-dependent Bayesian regret bound. This offers insight into how pretraining can greatly improve online performance and how the degree of improvement increases with the expert's competence level. We also develop a practical, approximate informed TS algorithm through Bayesian bootstrapping and show substantial empirical regret reduction through experiments.

Sample Efficient Deep Reinforcement Learning via Local Planning

The focus of this work is sample-efficient deep reinforcement learning (RL) with a simulator. One useful property of simulators is that it is typically easy to reset the environment to a previously observed state. We propose an algorithmic framework, named uncertainty-first local planning (UFLP), that takes advantage of this property. Concretely, in each data collection iteration, with some probability, our meta-algorithm resets the environment to an observed state which has high uncertainty, instead of sampling according to the initial-state distribution. The agent-environment interaction then proceeds as in the standard online RL setting. We demonstrate that this simple procedure can dramatically improve the sample cost of several baseline RL algorithms on difficult exploration tasks. Notably, with our framework, we can achieve super-human performance on the notoriously hard Atari game, Montezuma's Revenge, with a simple (distributional) double DQN. Our work can be seen as an efficient approximate implementation of an existing algorithm with theoretical guarantees, which offers an interpretation of the positive empirical results.

Regret Bounds for Information-Directed Reinforcement Learning

Information-directed sampling (IDS) has revealed its potential as a data-efficient algorithm for reinforcement learning (RL). However, theoretical understanding of IDS for Markov Decision Processes (MDPs) is still limited. We develop novel information-theoretic tools to bound the information ratio and cumulative information gain about the learning target. Our theoretical results shed light on the importance of choosing the learning target such that the practitioners can balance the computation and regret bounds. As a consequence, we derive prior-free Bayesian regret bounds for vanilla-IDS which learns the whole environment under tabular finite-horizon MDPs. In addition, we propose a computationally-efficient regularized-IDS that maximizes an additive form rather than the ratio form and show that it enjoys the same regret bound as vanilla-IDS. With the aid of rate-distortion theory, we improve the regret bound by learning a surrogate, less informative environment. Furthermore, we extend our analysis to linear MDPs and prove similar regret bounds for Thompson sampling as a by-product.

Contextual Information-Directed Sampling

Information-directed sampling (IDS) has recently demonstrated its potential as a data-efficient reinforcement learning algorithm. However, it is still unclear what is the right form of information ratio to optimize when contextual information is available. We investigate the IDS design through two contextual bandit problems: contextual bandits with graph feedback and sparse linear contextual bandits. We provably demonstrate the advantage of contextual IDS over conditional IDS and emphasize the importance of considering the context distribution. The main message is that an intelligent agent should invest more on the actions that are beneficial for the future unseen contexts while the conditional IDS can be myopic. We further propose a computationally-efficient version of contextual IDS based on Actor-Critic and evaluate it empirically on a neural network contextual bandit.

Interacting Contour Stochastic Gradient Langevin Dynamics

We propose an interacting contour stochastic gradient Langevin dynamics (ICSGLD) sampler, an embarrassingly parallel multiple-chain contour stochastic gradient Langevin dynamics (CSGLD) sampler with efficient interactions. We show that ICSGLD can be theoretically more efficient than a single-chain CSGLD with an equivalent computational budget. We also present a novel random-field function, which facilitates the estimation of self-adapting parameters in big data and obtains free mode explorations. Empirically, we compare the proposed algorithm with popular benchmark methods for posterior sampling. The numerical results show a great potential of ICSGLD for large-scale uncertainty estimation tasks.