We introduce the "Belief State Transformer", a next-token predictor that takes both a prefix and suffix as inputs, with a novel objective of predicting both the next token for the prefix and the previous token for the suffix. The Belief State Transformer effectively learns to solve challenging problems that conventional forward-only transformers struggle with, in a domain-independent fashion. Key to this success is learning a compact belief state that captures all relevant information necessary for accurate predictions. Empirical ablations show that each component of the model is essential in difficult scenarios where standard Transformers fall short. For the task of story writing with known prefixes and suffixes, our approach outperforms the Fill-in-the-Middle method for reaching known goals and demonstrates improved performance even when the goals are unknown. Altogether, the Belief State Transformer enables more efficient goal-conditioned decoding, better test-time inference, and high-quality text representations on small scale problems. Website: https://edwhu.github.io/bst-website

Pre-training on Internet data has proven to be a key ingredient for broad generalization in many modern ML systems. What would it take to enable such capabilities in robotic reinforcement learning (RL)? Offline RL methods, which learn from datasets of robot experience, offer one way to leverage prior data into the robotic learning pipeline. However, these methods have a "type mismatch" with video data (such as Ego4D), which are the largest prior datasets available for robotics, since video offers observation-only experience without the action or reward annotations needed for RL methods. In this paper, we develop a system for leveraging large-scale human video datasets in robotic offline RL, based entirely on learning value functions via temporal-difference learning. We show that value learning on video datasets learns representations that are more conducive to downstream robotic offline RL than other approaches for learning from video data. Our system, called V-PTR, combines the benefits of pre-training on video data with robotic offline RL approaches that train on diverse robot data, resulting in value functions and policies for manipulation tasks that perform better, act robustly, and generalize broadly. On several manipulation tasks on a real WidowX robot and in simulated settings, our framework produces policies that greatly improve over other prior methods. Our video and additional details can be found at https://dibyaghosh.com/vptr/.

Several self-supervised representation learning methods have been proposed for reinforcement learning (RL) with rich observations. For real world applications of RL, recovering underlying latent states is crucial, particularly when sensory inputs can contain irrelevant and exogenous information. In this work, we study how information bottlenecjs can be used to construct latent states efficiently in the presence of task irrelevant information. We propose architectures that utilize variational and discrete information bottleneck, coined as RepDIB, to learn structured factorized representations. Exploiting the expressiveness bought by factorized representations, we introduce a simple, yet effective, bottleneck that can be integrated with any existing self supervised objective for RL. We demonstrate this across several online and offline RL benchmarks, along with a real robot arm task, where we find that compressed representations with RepDIB can lead to strong performance improvements, as the learnt bottlenecks can help predict only the relevant state, while ignoring irrelevant information.

Learning to control an agent from offline data collected in a rich pixel-based visual observation space is vital for real-world applications of reinforcement learning (RL). A major challenge in this setting is the presence of input information that is hard to model and irrelevant to controlling the agent. This problem has been approached by the theoretical RL community through the lens of *exogenous information*, i.e., any control-irrelevant information contained in observations. For example, a robot navigating in busy streets needs to ignore irrelevant information, such as other people walking in the background, textures of objects, or birds in the sky. In this paper, we focus on the setting with visually detailed exogenous information and introduce new offline RL benchmarks that offer the ability to study this problem. We find that contemporary representation learning techniques can fail on datasets where the noise is a complex and time-dependent process, which is prevalent in practical applications. To address these, we propose to use multi-step inverse models to learn Agent-Centric Representations for Offline-RL (ACRO). Despite being simple and reward-free, we show theoretically and empirically that the representation created by this objective greatly outperforms baselines.

Informational parsimony provides a useful inductive bias for learning representations that achieve better generalization by being robust to noise and spurious correlations. We propose *information gating* as a way to learn parsimonious representations that identify the minimal information required for a task. When gating information, we can learn to reveal as little information as possible so that a task remains solvable, or hide as little information as possible so that a task becomes unsolvable. We gate information using a differentiable parameterization of the signal-to-noise ratio, which can be applied to arbitrary values in a network, e.g., erasing pixels at the input layer or activations in some intermediate layer. When gating at the input layer, our models learn which visual cues matter for a given task. When gating intermediate layers, our models learn which activations are needed for subsequent stages of computation. We call our approach *InfoGating*. We apply InfoGating to various objectives such as multi-step forward and inverse dynamics models, Q-learning, and behavior cloning, highlighting how InfoGating can naturally help in discarding information not relevant for control. Results show that learning to identify and use minimal information can improve generalization in downstream tasks. Policies based on InfoGating are considerably more robust to irrelevant visual features, leading to improved pretraining and finetuning of RL models.

Multi-step greedy policies have been extensively used in model-based reinforcement learning (RL), both when a model of the environment is available (e.g., in the game of Go) and when it is learned. In this paper, we explore their benefits in model-free RL, when employed using multi-step dynamic programming algorithms: $\kappa$-Policy Iteration ($\kappa$-PI) and $\kappa$-Value Iteration ($\kappa$-VI). These methods iteratively compute the next policy ($\kappa$-PI) and value function ($\kappa$-VI) by solving a surrogate decision problem with a shaped reward and a smaller discount factor. We derive model-free RL algorithms based on $\kappa$-PI and $\kappa$-VI in which the surrogate problem can be solved by any discrete or continuous action RL method, such as DQN and TRPO. We identify the importance of a hyper-parameter that controls the extent to which the surrogate problem is solved and suggest a way to set this parameter. When evaluated on a range of Atari and MuJoCo benchmark tasks, our results indicate that for the right range of $\kappa$, our algorithms outperform DQN and TRPO. This shows that our multi-step greedy algorithms are general enough to be applied over any existing RL algorithm and can significantly improve its performance.

Generating a curriculum for guided learning involves subjecting the agent to easier goals first, and then gradually increasing their difficulty. This work takes a similar direction and proposes a dual curriculum scheme for solving robotic manipulation tasks with sparse rewards, called MaMiC. It includes a macro curriculum scheme which divides the task into multiple subtasks followed by a micro curriculum scheme which enables the agent to learn between such discovered subtasks. We show how combining macro and micro curriculum strategies help in overcoming major exploratory constraints considered in robot manipulation tasks without having to engineer any complex rewards and also illustrate the meaning and usage of the individual curricula. The performance of such a scheme is analysed on the Fetch environments.

The options framework in reinforcement learning models the notion of a skill or a temporally extended sequence of actions. The discovery of a reusable set of skills has typically entailed building options, that navigate to bottleneck states. In this work, we instead adopt a complementary approach, where we attempt to discover options that navigate to landmark states. These states are prototypical representatives of well-connected regions and can hence access the associated region with relative ease. In this work, we propose Successor Options, which leverages Successor representations to build a model of the state space. The intra-option policies are learnt using a novel pseudo-reward and the model scales to high-dimensional spaces since it does not construct an explicit graph of the entire state space. Additionally, we also propose an Incremental Successor Options model that iterates between constructing Successor representations and building options, which is useful when robust Successor representations cannot be built solely from primitive actions. We demonstrate the efficacy of our approach on a collection of grid-worlds, and on the high-dimensional robotic control environment of Fetch.