A core challenge of Monte Carlo Tree Search (MCTS) is its sample efficiency, which can be improved by grouping state-action pairs and using their aggregate statistics instead of single-node statistics. On the Go Abstractions in Upper Confidence bounds applied to Trees (OGA-UCT) is the state-of-the-art MCTS abstraction algorithm for deterministic environments that builds its abstraction using the Abstractions of State-Action Pairs (ASAP) framework, which aims to detect states and state-action pairs with the same value under optimal play by analysing the search graph. ASAP, however, requires two state-action pairs to have the same immediate reward, which is a rigid condition that limits the number of abstractions that can be found and thereby the sample efficiency. In this paper, we break with the paradigm of grouping value-equivalent states or state-action pairs and instead group states and state-action pairs with possibly different values as long as the difference between their values can be inferred. We call this abstraction framework Known Value Difference Abstractions (KVDA), which infers the value differences by analysis of the immediate rewards and modifies OGA-UCT to use this framework instead. The modification is called KVDA-UCT, which detects significantly more abstractions than OGA-UCT, introduces no additional parameter, and outperforms OGA-UCT on a variety of deterministic environments and parameter settings.

Additionally, methods that utilize the bisimulation metric for evaluating state discrepancies face a theory-practice gap due to improper approximations in metric learning, particularly struggling with hard exploration tasks.

Assessing Large Language Models (LLMs)' underlying value differences enables comprehensive comparison of their misalignment, cultural adaptability, and biases. Nevertheless, current value measurement datasets face the informativeness challenge: with often outdated, contaminated, or generic test questions, they can only capture the shared value orientations among different LLMs, leading to saturated and thus uninformative results. To address this problem, we introduce AdAEM, a novel, self-extensible assessment framework for revealing LLMs' inclinations. Distinct from previous static benchmarks, AdAEM can automatically and adaptively generate and extend its test questions. This is achieved by probing the internal value boundaries of a diverse set of LLMs developed across cultures and time periods in an in-context optimization manner. The optimization process theoretically maximizes an information-theoretic objective to extract the latest or culturally controversial topics, providing more distinguishable and informative insights about models' value differences. In this way, AdAEM is able to co-evolve with the development of LLMs, consistently tracking their value dynamics. Using AdAEM, we generate 12,310 questions grounded in Schwartz Value Theory, conduct an extensive analysis to manifest our method's validity and effectiveness, and benchmark the values of 16 LLMs, laying the groundwork for better value research.

We address the challenge of quantifying Bayesian uncertainty and incorporating it in offline use cases of finite-state Markov Decision Processes (MDPs) with unknown dynamics. Our approach provides a principled method to disentangle epistemic and aleatoric uncertainty, and a novel technique to find policies that optimise Bayesian posterior expected value without relying on strong assumptions about the MDP's posterior distribution. First, we utilise standard Bayesian reinforcement learning methods to capture the posterior uncertainty in MDP parameters based on available data. We then analytically compute the first two moments of the return distribution across posterior samples and apply the law of total variance to disentangle aleatoric and epistemic uncertainties. To find policies that maximise posterior expected value, we leverage the closed-form expression for value as a function of policy. This allows us to propose a stochastic gradient-based approach for solving the problem. We illustrate the uncertainty quantification and Bayesian posterior value optimisation performance of our agent in simple, interpretable gridworlds and validate it through ground-truth evaluations on synthetic MDPs. Finally, we highlight the real-world impact and computational scalability of our method by applying it to the AI Clinician problem, which recommends treatment for patients in intensive care units and has emerged as a key use case of finite-state MDPs with offline data. We discuss the challenges that arise with Bayesian modelling of larger scale MDPs while demonstrating the potential to apply our methods rooted in Bayesian decision theory into the real world. We make our code available at https://github.com/filippovaldettaro/finite-state-mdps .

There is a growing interest in estimating heterogeneous treatment effects across individuals using their high-dimensional feature attributes. Achieving high performance in such high-dimensional heterogeneous treatment effect estimation is challenging because in this setup, it is usual that some features induce sample selection bias while others do not but are predictive of potential outcomes. To avoid losing such predictive feature information, existing methods learn separate feature representations using inverse probability weighting (IPW). However, due to their numerically unstable IPW weights, these methods suffer from estimation bias under a finite sample setup. To develop a numerically robust estimator by weighted representation learning, we propose a differentiable Pareto-smoothed weighting framework that replaces extreme weight values in an end-to-end fashion. Our experimental results show that by effectively correcting the weight values, our proposed method outperforms the existing ones, including traditional weighting schemes. Our code is available at https://github.com/ychika/DPSW.

We present a probabilistic algorithm for nonlinear inverse reinforcement learning. The goal of inverse reinforcement learning is to learn the reward function in a Markov decision process from expert demonstrations. While most prior inverse reinforcement learning algorithms represent the reward as a linear combination of a set of features, we use Gaussian processes to learn the reward as a nonlinear function, while also determining the relevance of each feature to the expert's policy. Our probabilistic algorithm allows complex behaviors to be captured from suboptimal stochastic demonstrations, while automatically balancing the simplicity of the learned reward structure against its consistency with the observed actions.

We consider the problem of policy transfer between two Markov Decision Processes (MDPs). We introduce a lemma based on existing theoretical results in reinforcement learning to measure the relativity gap between two arbitrary MDPs, that is the difference between any two cumulative expected returns defined on different policies and environment dynamics. Based on this lemma, we propose two new algorithms referred to as Relative Policy Optimization (RPO) and Relative Transition Optimization (RTO), which offer fast policy transfer and dynamics modelling, respectively. RPO transfers the policy evaluated in one environment to maximize the return in another, while RTO updates the parameterized dynamics model to reduce the gap between the dynamics of the two environments. Integrating the two algorithms results in the complete Relative Policy-Transition Optimization (RPTO) algorithm, in which the policy interacts with the two environments simultaneously, such that data collections from two environments, policy and transition updates are completed in one closed loop to form a principled learning framework for policy transfer. We demonstrate the effectiveness of RPTO on a set of MuJoCo continuous control tasks by creating policy transfer problems via variant dynamics.

This paper addresses the problem of inverse reinforcement learning (IRL) - inferring the reward function of an agent from observing its behavior. IRL can provide a generalizable and compact representation for apprenticeship learning, and enable accurately inferring the preferences of a human in order to assist them. However, effective IRL is challenging, because many reward functions can be compatible with an observed behavior. We focus on how prior reinforcement learning (RL) experience can be leveraged to make learning these preferences faster and more efficient. We propose the IRL algorithm BASIS (Behavior Acquisition through Successor-feature Intention inference from Samples), which leverages multi-task RL pre-training and successor features to allow an agent to build a strong basis for intentions that spans the space of possible goals in a given domain. When exposed to just a few expert demonstrations optimizing a novel goal, the agent uses its basis to quickly and effectively infer the reward function. Our experiments reveal that our method is highly effective at inferring and optimizing demonstrated reward functions, accurately inferring reward functions from less than 100 trajectories. Inverse reinforcement learning (IRL) seeks to identify a reward function under which observed behavior of an expert is optimal. Once an agent has effectively inferred the reward function, it can then use standard (forward) RL to optimize it, and thus acquire not only useful skills by observing demonstrations, but also a reward function as an explanation for the demonstrator's behavior. By inferring the underlying goal being pursued by the demonstrator, the agent is more likely to be able to generalize to a new scenario in which it must optimize that goal, versus an agent which merely imitates the demonstrated actions. IRL has already proven useful in applications including autonomous driving, where learned models capture the behavior of nearby drivers and pedestrians (Huang et al., 2021; Kim & Pineau, 2016), and is a key component in enabling assistive technologies where a helper agent must infer the goals of the human it is assisting (Hadfield-Menell et al., 2016).

Meta reinforcement learning (meta-RL) aims to learn a policy solving a set of training tasks simultaneously and quickly adapting to new tasks. It requires massive amounts of data drawn from training tasks to infer the common structure shared among tasks. Without heavy reward engineering, the sparse rewards in long-horizon tasks exacerbate the problem of sample efficiency in meta-RL. Another challenge in meta-RL is the discrepancy of difficulty level among tasks, which might cause one easy task dominating learning of the shared policy and thus preclude policy adaptation to new tasks. This work introduces a novel objective function to learn an action translator among training tasks. We theoretically verify that the value of the transferred policy with the action translator can be close to the value of the source policy and our objective function (approximately) upper bounds the value difference. We propose to combine the action translator with context-based meta-RL algorithms for better data collection and more efficient exploration during meta-training. Our approach empirically improves the sample efficiency and performance of meta-RL algorithms on sparse-reward tasks.