NeurIPS 2024 Papers

Offline policy evaluation (OPE) allows us to evaluate and estimate a new sequential decision-making policy's performance by leveraging historical interaction data collected from other policies. Evaluating a new policy online without a confident estimate of its performance can lead to costly, unsafe, or hazardous outcomes, especially in education and healthcare. Several OPE estimators have been proposed in the last decade, many of which have hyperparameters and require training. Unfortunately, choosing the best OPE algorithm for each task and domain is still unclear. In this paper, we propose a new algorithm that adaptively blends a set of OPE estimators given a dataset without relying on an explicit selection using a statistical procedure. We prove that our estimator is consistent and satisfies several desirable properties for policy evaluation. Additionally, we demonstrate that when compared to alternative approaches, our estimator can be used to select higher-performing policies in healthcare and robotics. Our work contributes to improving ease of use for a general-purpose, estimator-agnostic, off-policy evaluation framework for offline RL.

Despite advancements in robotic-assisted surgery, automating complex tasks like suturing remains challenging due to the need for adaptability and precision. Learning-based approaches, particularly reinforcement learning (RL) and imitation learning (IL), require realistic simulation environments for efficient data collection. However, current platforms often include only relatively simple, non-dexterous manipulations and lack the flexibility required for effective learning and generalization. We introduce SurgicAI, a novel platform for development and benchmarking that addresses these challenges by providing the flexibility to accommodate both modular subtasks and more importantly task decomposition in RL-based surgical robotics. Compatible with the da Vinci Surgical System, SurgicAI offers a standardized pipeline for collecting and utilizing expert demonstrations. It supports the deployment of multiple RL and IL approaches, and the training of both singular and compositional subtasks in suturing scenarios, featuring high dexterity and modularization. Meanwhile, SurgicAI sets clear metrics and benchmarks for the assessment of learned policies. We implemented and evaluated multiple RL and IL algorithms on SurgicAI. Our detailed benchmark analysis underscores SurgicAI's potential to advance policy learning in surgical robotics. Details: https://github.com/surgical-robotics-ai/SurgicAI

In this paper, we study the non-asymptotic sample complexity for the pure exploration problem in contextual bandits and tabular reinforcement learning (RL): identifying an $\epsilon$-optimal policy from a set of policies $\Pi$ with high probability. Existing work in bandits has shown that it is possible to identify the best policy by estimating only the *difference* between the behaviors of individual policies–which can have substantially lower variance than estimating the behavior of each policy directly—yet the best-known complexities in RL fail to take advantage of this, and instead estimate the behavior of each policy directly. Does it suffice to estimate only the differences in the behaviors of policies in RL? We answer this question positively for contextual bandits, but in the negative for tabular RL, showing a separation between contextual bandits and RL. However, inspired by this, we show that it *almost* suffices to estimate only the differences in RL: if we can estimate the behavior of a *single* reference policy, it suffices to only estimate how any other policy deviates from this reference policy. We develop an algorithm which instantiates this principle and obtains, to the best of our knowledge, the tightest known bound on the sample complexity of tabular RL.

Traditional 3D shape reconstruction techniques from multi-view images, such as structure from motion and multi-view stereo, face challenges in reconstructing transparent objects. Recent advances in neural radiance fields and its variants primarily address opaque or transparent objects, encountering difficulties to reconstruct both transparent and opaque objects simultaneously. This paper introduces $\alpha$-NeuS$\textemdash$an extension of NeuS$\textemdash$that proves NeuS is unbiased for materials from fully transparent to fully opaque. We find that transparent and opaque surfaces align with the non-negative local minima and the zero iso-surface, respectively, in the learned distance field of NeuS. Traditional iso-surfacing extraction algorithms, such as marching cubes, which rely on fixed iso-values, are ill-suited for such data. We develop a method to extract the transparent and opaque surface simultaneously based on DCUDF. To validate our approach, we construct a benchmark that includes both real-world and synthetic scenes, demonstrating its practical utility and effectiveness. Our data and code are publicly available at https://github.com/728388808/alpha-NeuS.

Occupancy functions play an instrumental role in reinforcement learning (RL) for guiding exploration, handling distribution shift, and optimizing general objectives beyond the expected return. Yet, computationally efficient policy optimization methods that use (only) occupancy functions are virtually non-existent. In this paper, we establish the theoretical foundations of model-free policy gradient (PG) methods that compute the gradient through the occupancy for both online and offline RL, without modeling value functions. Our algorithms reduce gradient estimation to squared-loss regression and are computationally oracle-efficient. We characterize the sample complexities of both local and global convergence, accounting for both finite-sample estimation error and the roles of exploration (online) and data coverage (offline). Occupancy-based PG naturally handles arbitrary offline data distributions, and, with one-line algorithmic changes, can be adapted to optimize any differentiable objective functional.

Continual learning with deep neural networks presents challenges distinct from both the fixed-dataset and convex continual learning regimes. One such challenge is plasticity loss, wherein a neural network trained in an online fashion displays a degraded ability to fit new tasks. This problem has been extensively studied in both supervised learning and off-policy reinforcement learning (RL), where a number of remedies have been proposed. Still, plasticity loss has received less attention in the on-policy deep RL setting. Here we perform an extensive set of experiments examining plasticity loss and a variety of mitigation methods in on-policy deep RL. We demonstrate that plasticity loss is pervasive under domain shift in this regime, and that a number of methods developed to resolve it in other settings fail, sometimes even performing worse than applying no intervention at all. In contrast, we find that a class of ``regenerative'' methods are able to consistently mitigate plasticity loss in a variety of contexts, including in gridworld tasks and more challenging environments like Montezuma's Revenge and ProcGen.

With high-dimensional state spaces, visual reinforcement learning (RL) faces significant challenges in exploitation and exploration, resulting in low sample efficiency and training stability. As a time-efficient diffusion model, although consistency models have been validated in online state-based RL, it is still an open question whether it can be extended to visual RL. In this paper, we investigate the impact of non-stationary distribution and the actor-critic framework on consistency policy in online RL, and find that consistency policy was unstable during the training, especially in visual RL with the high-dimensional state space. To this end, we suggest sample-based entropy regularization to stabilize the policy training, and propose a consistency policy with prioritized proximal experience regularization (CP3ER) to improve sample efficiency. CP3ER achieves new state-of-the-art (SOTA) performance in 21 tasks across DeepMind control suite and Meta-world. To our knowledge, CP3ER is the first method to apply diffusion/consistency models to visual RL and demonstrates the potential of consistency models in visual RL.

Federated Learning (FL) allows multiple clients to collaboratively train models without directly sharing their private data. While various data augmentation techniques have been actively studied in the FL environment, most of these methods share input-level or feature-level data information over communication, posing potential privacy leakage. In response to this challenge, we introduce a federated data augmentation algorithm named FedAvP that shares only the augmentation policies, not the data-related information. For data security and efficient policy search, we interpret the policy loss as a meta update loss in standard FL algorithms and utilize the first-order gradient information to further enhance privacy and reduce communication costs. Moreover, we propose a meta-learning method to search for adaptive personalized policies tailored to heterogeneous clients. Our approach outperforms existing best performing augmentation policy search methods and federated data augmentation methods, in the benchmarks for heterogeneous FL.

LLMs show remarkable emergent abilities, such as inferring concepts from presumably out-of-distribution prompts, known as in-context learning. Though this success is often attributed to the Transformer architecture, our systematic understanding is limited. In complex real-world data sets, even defining what is out-of-distribution is not obvious. To better understand the OOD behaviour of autoregressive LLMs, we focus on formal languages, which are defined by the intersection of rules. We define a new scenario of OOD compositional generalization, termed \textit{rule extrapolation}. Rule extrapolation describes OOD scenarios, where the prompt violates at least one rule. We evaluate rule extrapolation in formal languages with varying complexity in linear and recurrent architectures, the Transformer, and state space models to understand the architectures' influence on rule extrapolation. We also lay the first stones of a normative theory of rule extrapolation, inspired by the Solomonoff prior in algorithmic information theory.

Reward specification is one of the most tricky problems in Reinforcement Learning, which usually requires tedious hand engineering in practice. One promising approach to tackle this challenge is to adopt existing expert video demonstrations for policy learning. Some recent work investigates how to learn robot policies from only a single/few expert video demonstrations. For example, reward labeling via Optimal Transport (OT) has been shown to be an effective strategy to generate a proxy reward by measuring the alignment between the robot trajectory and the expert demonstrations. However, previous work mostly overlooks that the OT reward is invariant to temporal order information, which could bring extra noise to the reward signal. To address this issue, in this paper, we introduce the Temporal Optimal Transport (TemporalOT) reward to incorporate temporal order information for learning a more accurate OT-based proxy reward. Extensive experiments on the Meta-world benchmark tasks validate the efficacy of the proposed method. Our code is available at: https://github.com/fuyw/TemporalOT.

This paper studies Generative Flow Networks (GFlowNets), which learn to sample objects proportionally to a given reward function through the trajectory of state transitions. In this work, we observe that GFlowNets tend to under-exploit the high-reward objects due to training on insufficient number of trajectories, which may lead to a large gap between the estimated flow and the (known) reward value. In response to this challenge, we propose a pessimistic backward policy for GFlowNets (PBP-GFN), which maximizes the observed flow to align closely with the true reward for the object. We extensively evaluate PBP-GFN across eight benchmarks, including hyper-grid environment, bag generation, structured set generation, molecular generation, and four RNA sequence generation tasks. In particular, PBP-GFN enhances the discovery of high-reward objects, maintains the diversity of the objects, and consistently outperforms existing methods.

Existing Goal-Conditioned Reinforcement Learning (GCRL) algorithms are built upon Hindsight Experience Replay (HER), which densifies rewards through hindsight replay and leverages historical goal-achieving information to construct a learning curriculum. However, when the task is characterized by a non-Markovian reward (NMR), whose computation depends on multiple steps of states and actions, HER can no longer densify rewards by treating a single encountered state as the hindsight goal. The lack of informative rewards hinders policy learning, resulting in rolling out failed trajectories. Consequently, the replay buffer is overwhelmed with failed trajectories, impeding the establishment of an applicable curriculum. To circumvent these limitations, we deviate from existing HER-based methods and propose an on-policy GCRL framework, GCPO, which is applicable to both multi-goal Markovian reward (MR) and NMR problems. GCPO consists of (1) Pre-training from Demonstrations, which pre-trains the policy to possess an initial goal-achieving capability, thereby diminishing the difficulty of subsequent online learning. (2) Online Self-Curriculum Learning, which first estimates the policy's goal-achieving capability based on historical evaluation information and then selects progressively challenging goals for learning based on its current capability. We evaluate GCPO on a challenging multi-goal long-horizon task: fixed-wing UAV velocity vector control. Experimental results demonstrate that GCPO is capable of effectively addressing both multi-goal MR and NMR problems.

Tokenization is widely used in large language models because it significantly improves performance. However, tokenization imposes several disadvantages, such as performance biases, increased adversarial vulnerability, decreased character-level modeling performance, and increased modeling complexity. To address these disadvantages without sacrificing performance, we propose SpaceByte, a novel byte-level decoder architecture that closes the performance gap between byte-level and subword autoregressive language modeling. SpaceByte consists of a byte-level Transformer model, but with extra larger transformer blocks inserted in the middle of the layers. We find that performance is significantly improved by applying these larger blocks only after certain bytes, such as space characters, which typically denote word boundaries. Our experiments show that for a fixed training and inference compute budget, SpaceByte outperforms other byte-level architectures and roughly matches the performance of tokenized Transformer architectures.

Despite Flow Matching and diffusion models having emerged as powerful generative paradigms for continuous variables such as images and videos, their application to high-dimensional discrete data, such as language, is still limited. In this work, we present Discrete Flow Matching, a novel discrete flow paradigm designed specifically for generating discrete data. Discrete Flow Matching offers several key contributions: (i) it works with a general family of probability paths interpolating between source and target distributions; (ii) it allows for a generic formula for sampling from these probability paths using learned posteriors such as the probability denoiser ($x$-prediction) and noise-prediction ($\epsilon$-prediction); (iii) practically, focusing on specific probability paths defined with different schedulers improves generative perplexity compared to previous discrete diffusion and flow models; and (iv) by scaling Discrete Flow Matching models up to 1.7B parameters, we reach 6.7% Pass@1 and 13.4% Pass@10 on HumanEval and 6.7% Pass@1 and 20.6% Pass@10 on 1-shot MBPP coding benchmarks. Our approach is capable of generating high-quality discrete data in a non-autoregressive fashion, significantly closing the gap between autoregressive models and discrete flow models.

We introduce Language Feedback Models (LFMs) that identify desirable behaviour --- actions that help achieve tasks specified in the instruction - for imitation learning in instruction following. To train LFMs, we obtain feedback from Large Language Models (LLMs) on visual trajectories verbalized to language descriptions. First, by using LFMs to identify desirable behaviour to imitate, we improve in task-completion rate over strong behavioural cloning baselines on three distinct language grounding environments (Touchdown, ScienceWorld, and ALFWorld). Second, LFMs outperform using LLMs as experts to directly predict actions, when controlling for the number of LLM output tokens. Third, LFMs generalize to unseen environments, improving task-completion rate by 3.5-12.0% through one round of adaptation. Finally, LFMs can be modified to provide human-interpretable feedback without performance loss, allowing human verification of desirable behaviour for imitation learning.

Modern deep policy gradient methods achieve effective performance on simulated robotic tasks, but they all require large replay buffers or expensive batch updates, or both, making them incompatible for real systems with resource-limited computers. We show that these methods fail catastrophically when limited to small replay buffers or during *incremental learning*, where updates only use the most recent sample without batch updates or a replay buffer. We propose a novel incremental deep policy gradient method --- *Action Value Gradient (AVG)* and a set of normalization and scaling techniques to address the challenges of instability in incremental learning. On robotic simulation benchmarks, we show that AVG is the only incremental method that learns effectively, often achieving final performance comparable to batch policy gradient methods. This advancement enabled us to show for the first time effective deep reinforcement learning with real robots using only incremental updates, employing a robotic manipulator and a mobile robot.

Large language models (LLMs) are increasingly essential in processing natural languages, yet their application is frequently compromised by biases and inaccuracies originating in their training data. In this study, we introduce \textbf{Cross-Care}, the first benchmark framework dedicated to assessing biases and real world knowledge in LLMs, specifically focusing on the representation of disease prevalence across diverse demographic groups. We systematically evaluate how demographic biases embedded in pre-training corpora like $ThePile$ influence the outputs of LLMs. We expose and quantify discrepancies by juxtaposing these biases against actual disease prevalences in various U.S. demographic groups. Our results highlight substantial misalignment between LLM representation of disease prevalence and real disease prevalence rates across demographic subgroups, indicating a pronounced risk of bias propagation and a lack of real-world grounding for medical applications of LLMs. Furthermore, we observe that various alignment methods minimally resolve inconsistencies in the models' representation of disease prevalence across different languages. For further exploration and analysis, we make all data and a data visualization tool available at: \url{www.crosscare.net}.

We study the batched best arm identification (BBAI) problem, where the learner's goal is to identify the best arm while switching the policy as less as possible. In particular, we aim to find the best arm with probability $1-\delta$ for some small constant $\delta>0$ while minimizing both the sample complexity (total number of arm pulls) and the batch complexity (total number of batches). We propose the three-batch best arm identification (Tri-BBAI) algorithm, which is the first batched algorithm that achieves the optimal sample complexity in the asymptotic setting (i.e., $\delta\rightarrow 0$) and runs in $3$ batches in expectation. Based on Tri-BBAI, we further propose the almost optimal batched best arm identification (Opt-BBAI) algorithm, which is the first algorithm that achieves the near-optimal sample and batch complexity in the non-asymptotic setting (i.e., $1/\delta$ is finite), while enjoying the same batch and sample complexity as Tri-BBAI when $\delta$ tends to zero. Moreover, in the non-asymptotic setting, the complexity of previous batch algorithms is usually conditioned on the event that the best arm is returned (with a probability of at least $1-\delta$), which is potentially unbounded in cases where a sub-optimal arm is returned. In contrast, the complexity of Opt-BBAI does not rely on such an event. This is achieved through a novel procedure that we design for checking whether the best arm is eliminated, which is of independent interest.