Search

Diffusion models have demonstrated their capabilities in modeling trajectories of multi-tasks. However, existing multi-task planners or policies typically rely on task-specific demonstrations via multi-task imitation, or require task-specific reward labels to facilitate policy optimization via Reinforcement Learning (RL). To address these challenges, we aim to develop a versatile diffusion planner that can leverage large-scale inferior data that contains task-agnostic sub-optimal trajectories, with the ability to fast adapt to specific tasks. In this paper, we propose \textbf{SODP}, a two-stage framework that leverages \textbf{S}ub-\textbf{O}ptimal data to learn a \textbf{D}iffusion \textbf{P}lanner, which is generalizable for various downstream tasks. Specifically, in the pre-training stage, we train a foundation diffusion planner that extracts general planning capabilities by modeling the versatile distribution of multi-task trajectories, which can be sub-optimal and has wide data coverage. Then for downstream tasks, we adopt RL-based fine-tuning with task-specific rewards to fast refine the diffusion planner, which aims to generate action sequences with higher task-specific returns. Experimental results from multi-task domains including Meta-World and Adroit demonstrate that SODP outperforms state-of-the-art methods with only a small amount of data for reward-guided fine-tuning.

Stochastic diffusion processes are pervasive in nature, from the seemingly erratic Brownian motion to the complex interactions of synaptically-coupled spiking neurons. Recently, drawing inspiration from Langevin dynamics, neuromorphic diffusion models were proposed and have become one of the major breakthroughs in the field of generative artificial intelligence. Unlike discriminative models that have been well developed to tackle classification or regression tasks, diffusion models as well as other generative models such as ChatGPT aim at creating content based upon contexts learned. However, the more complex algorithms of these models result in high computational costs using today's technologies, creating a bottleneck in their efficiency, and impeding further development. Here, we develop a spintronic voltage-controlled magnetoelectric memory hardware for the neuromorphic diffusion process. The in-memory computing capability of our spintronic devices goes beyond current Von Neumann architecture, where memory and computing units are separated. Together with the non-volatility of magnetic memory, we can achieve high-speed and low-cost computing, which is desirable for the increasing scale of generative models in the current era. We experimentally demonstrate that the hardware-based true random diffusion process can be implemented for image generation and achieve comparable image quality to software-based training as measured by the Frechet inception distance (FID) score, achieving ~10^3 better energy-per-bit-per-area over traditional hardware.

The Generative Flow Network (GFlowNet) is a probabilistic framework in which an agent learns a stochastic policy and flow functions to sample objects with probability proportional to an unnormalized reward function. GFlowNets share a strong resemblance to reinforcement learning (RL), that typically aims to maximize reward, due to their sequential decision-making processes. Recent works have studied connections between GFlowNets and maximum entropy (MaxEnt) RL, which modifies the standard objective of RL agents by learning an entropy-regularized objective. However, a critical theoretical gap persists: despite the apparent similarities in their sequential decision-making nature, a direct link between GFlowNets and standard RL has yet to be discovered, while bridging this gap could further unlock the potential of both fields. In this paper, we establish a new connection between GFlowNets and policy evaluation for a uniform policy. Surprisingly, we find that the resulting value function for the uniform policy has a close relationship to the flows in GFlowNets. Leveraging these insights, we further propose a novel rectified policy evaluation (RPE) algorithm, which achieves the same reward-matching effect as GFlowNets, offering a new perspective. We compare RPE, MaxEnt RL, and GFlowNets in a number of benchmarks, and show that RPE achieves competitive results compared to previous approaches. This work sheds light on the previously unexplored connection between (non-MaxEnt) RL and GFlowNets, potentially opening new avenues for future research in both fields.

Generative Flow Networks (GFlowNets) are amortized sampling methods for learning a stochastic policy to sequentially generate compositional objects with probabilities proportional to their rewards. GFlowNets exhibit a remarkable ability to generate diverse sets of high-reward objects, in contrast to standard return maximization reinforcement learning approaches, which often converge to a single optimal solution. Recent works have arisen for learning goal-conditioned GFlowNets to acquire various useful properties, aiming to train a single GFlowNet capable of achieving different goals as the task specifies. However, training a goal-conditioned GFlowNet poses critical challenges due to extremely sparse rewards, which is further exacerbated in large state spaces. In this work, we propose a novel method named Retrospective Backward Synthesis (RBS) to address these challenges. Specifically, RBS synthesizes a new backward trajectory based on the backward policy in GFlowNets to enrich training trajectories with enhanced quality and diversity, thereby efficiently solving the sparse reward problem. Extensive empirical results show that our method improves sample efficiency by a large margin and outperforms strong baselines on various standard evaluation benchmarks.

Acquiring a multi-task imitation policy in 3D manipulation poses challenges in terms of scene understanding and action prediction. Current methods employ both 3D representation and multi-view 2D representation to predict the poses of the robot's end-effector. However, they still require a considerable amount of high-quality robot trajectories, and suffer from limited generalization in unseen tasks and inefficient execution in long-horizon reasoning. In this paper, we propose SAM-E, a novel architecture for robot manipulation by leveraging a vision-foundation model for generalizable scene understanding and sequence imitation for long-term action reasoning. Specifically, we adopt Segment Anything (SAM) pre-trained on a huge number of images and promptable masks as the foundation model for extracting task-relevant features, and employ parameter-efficient fine-tuning on robot data for a better understanding of embodied scenarios. To address long-horizon reasoning, we develop a novel multi-channel heatmap that enables the prediction of the action sequence in a single pass, notably enhancing execution efficiency. Experimental results from various instruction-following tasks demonstrate that SAM-E achieves superior performance with higher execution efficiency compared to the baselines, and also significantly improves generalization in few-shot adaptation to new tasks., Comment: ICML 2024. Project page: https://sam-embodied.github.io

Sequential decision-making is desired to align with human intents and exhibit versatility across various tasks. Previous methods formulate it as a conditional generation process, utilizing return-conditioned diffusion models to directly model trajectory distributions. Nevertheless, the return-conditioned paradigm relies on pre-defined reward functions, facing challenges when applied in multi-task settings characterized by varying reward functions (versatility) and showing limited controllability concerning human preferences (alignment). In this work, we adopt multi-task preferences as a unified condition for both single- and multi-task decision-making, and propose preference representations aligned with preference labels. The learned representations are used to guide the conditional generation process of diffusion models, and we introduce an auxiliary objective to maximize the mutual information between representations and corresponding generated trajectories, improving alignment between trajectories and preferences. Extensive experiments in D4RL and Meta-World demonstrate that our method presents favorable performance in single- and multi-task scenarios, and exhibits superior alignment with preferences., Comment: Accepted by NeurIPS 2024, 23 pages

Learning a generalist embodied agent capable of completing multiple tasks poses challenges, primarily stemming from the scarcity of action-labeled robotic datasets. In contrast, a vast amount of human videos exist, capturing intricate tasks and interactions with the physical world. Promising prospects arise for utilizing actionless human videos for pre-training and transferring the knowledge to facilitate robot policy learning through limited robot demonstrations. However, it remains a challenge due to the domain gap between humans and robots. Moreover, it is difficult to extract useful information representing the dynamic world from human videos, because of its noisy and multimodal data structure. In this paper, we introduce a novel framework to tackle these challenges, which leverages a unified discrete diffusion to combine generative pre-training on human videos and policy fine-tuning on a small number of action-labeled robot videos. We start by compressing both human and robot videos into unified video tokens. In the pre-training stage, we employ a discrete diffusion model with a mask-and-replace diffusion strategy to predict future video tokens in the latent space. In the fine-tuning stage, we harness the imagined future videos to guide low-level action learning with a limited set of robot data. Experiments demonstrate that our method generates high-fidelity future videos for planning and enhances the fine-tuned policies compared to previous state-of-the-art approaches with superior performance. Our project website is available at https://video-diff.github.io/., Comment: Accepted by NeurIPS 2024. 24 pages

Diffusion models surpass previous generative models in sample quality and training stability. Recent works have shown the advantages of diffusion models in improving reinforcement learning (RL) solutions. This survey aims to provide an overview of this emerging field and hopes to inspire new avenues of research. First, we examine several challenges encountered by RL algorithms. Then, we present a taxonomy of existing methods based on the roles of diffusion models in RL and explore how the preceding challenges are addressed. We further outline successful applications of diffusion models in various RL-related tasks. Finally, we conclude the survey and offer insights into future research directions. We are actively maintaining a GitHub repository for papers and other related resources in utilizing diffusion models in RL: https://github.com/apexrl/Diff4RLSurvey., Comment: Fixed typos

A quantum anomalous Hall (QAH) insulator breaks reciprocity by combining magnetic polarization and spin-orbit coupling to generate a unidirectional transmission of signals in the absence of an external magnetic field. Such behavior makes QAH materials a good platform for the innovation of circulator technologies. However, it remains elusive as to how the wavelength of the chiral edge plasmon relates to its frequency and how the plasmon wave packet is excited in the time domain in a QAH insulator. Here, we investigate the edge magnetoplasmon (EMP) resonances in Cr-(Bi,Sb)$_2$Te$_3$ by frequency and time domain measurements. From disk shaped samples with various dimensions, we obtain the dispersion relation of EMPs and extract the drift velocity of the chiral edge state. From the time-resolved transport measurements, we identify the velocity of the plasmon wave packet and observe a transition from the edge to bulk transport at an elevated temperature. We show that the frequency and time domain measurements are well modeled by loss from the microwave induced dissipative channels in the bulk area. Our results demonstrate that the EMP decay rate can be significantly reduced by applying a low microwave power and fabricating devices of larger diameter $\ge100~\mu$m. In a $R=125~\mu$m sample, a non-reciprocity of 20 dB has been realized at 1.3 GHz, shining light on using QAH insulators to develop on-chip non-reciprocal devices.

The robustness of legged locomotion is crucial for quadrupedal robots in challenging terrains. Recently, Reinforcement Learning (RL) has shown promising results in legged locomotion and various methods try to integrate privileged distillation, scene modeling, and external sensors to improve the generalization and robustness of locomotion policies. However, these methods are hard to handle uncertain scenarios such as abrupt terrain changes or unexpected external forces. In this paper, we consider a novel risk-sensitive perspective to enhance the robustness of legged locomotion. Specifically, we employ a distributional value function learned by quantile regression to model the aleatoric uncertainty of environments, and perform risk-averse policy learning by optimizing the worst-case scenarios via a risk distortion measure. Extensive experiments in both simulation environments and a real Aliengo robot demonstrate that our method is efficient in handling various external disturbances, and the resulting policy exhibits improved robustness in harsh and uncertain situations in legged locomotion. Videos are available at https://risk-averse-locomotion.github.io/., Comment: 8 pages, 5 figures

Topological insulators (TI) and magnetic topological insulators (MTI) can apply highly efficient spin-orbit torque (SOT) and manipulate the magnetization with their unique topological surface states with ultra-high efficiency. Here, we demonstrate efficient SOT switching of a hard MTI, V-doped (Bi,Sb)2Te3 (VBST) with a large coercive field that can prevent the influence of an external magnetic field. A giant switched anomalous Hall resistance of 9.2 $k\Omega$ is realized, among the largest of all SOT systems, which makes the Hall channel a good readout and eliminates the need to fabricate complicated magnetic tunnel junction (MTJ) structures. The SOT switching current density can be reduced to $2.8\times10^5 A/cm^2$. Moreover, as the Fermi level is moved away from the Dirac point by both gate and composition tuning, VBST exhibits a transition from edge-state-mediated to surface-state-mediated transport, thus enhancing the SOT effective field to $1.56\pm 0.12 T/ (10^6 A/cm^2)$ and the interfacial charge-to-spin conversion efficiency to $3.9\pm 0.3 nm^{-1}$. The findings establish VBST as an extraordinary candidate for energy-efficient magnetic memory devices.

Reinforcement Learning (RL) has recently achieved remarkable success in robotic control. However, most works in RL operate in simulated environments where privileged knowledge (e.g., dynamics, surroundings, terrains) is readily available. Conversely, in real-world scenarios, robot agents usually rely solely on local states (e.g., proprioceptive feedback of robot joints) to select actions, leading to a significant sim-to-real gap. Existing methods address this gap by either gradually reducing the reliance on privileged knowledge or performing a two-stage policy imitation. However, we argue that these methods are limited in their ability to fully leverage the available privileged knowledge, resulting in suboptimal performance. In this paper, we formulate the sim-to-real gap as an information bottleneck problem and therefore propose a novel privileged knowledge distillation method called the Historical Information Bottleneck (HIB). In particular, HIB learns a privileged knowledge representation from historical trajectories by capturing the underlying changeable dynamic information. Theoretical analysis shows that the learned privileged knowledge representation helps reduce the value discrepancy between the oracle and learned policies. Empirical experiments on both simulated and real-world tasks demonstrate that HIB yields improved generalizability compared to previous methods. Videos of real-world experiments are available at https://sites.google.com/view/history-ib ., Comment: Accepted by CoRL 2024

Diffusion models have demonstrated highly-expressive generative capabilities in vision and NLP. Recent studies in reinforcement learning (RL) have shown that diffusion models are also powerful in modeling complex policies or trajectories in offline datasets. However, these works have been limited to single-task settings where a generalist agent capable of addressing multi-task predicaments is absent. In this paper, we aim to investigate the effectiveness of a single diffusion model in modeling large-scale multi-task offline data, which can be challenging due to diverse and multimodal data distribution. Specifically, we propose Multi-Task Diffusion Model (\textsc{MTDiff}), a diffusion-based method that incorporates Transformer backbones and prompt learning for generative planning and data synthesis in multi-task offline settings. \textsc{MTDiff} leverages vast amounts of knowledge available in multi-task data and performs implicit knowledge sharing among tasks. For generative planning, we find \textsc{MTDiff} outperforms state-of-the-art algorithms across 50 tasks on Meta-World and 8 maps on Maze2D. For data synthesis, \textsc{MTDiff} generates high-quality data for testing tasks given a single demonstration as a prompt, which enhances the low-quality datasets for even unseen tasks., Comment: Accepted by NeurIPS 2023. 22 pages

Leveraging learned strategies in unfamiliar scenarios is fundamental to human intelligence. In reinforcement learning, rationally reusing the policies acquired from other tasks or human experts is critical for tackling problems that are difficult to learn from scratch. In this work, we present a framework called Selective Myopic bEhavior Control~(SMEC), which results from the insight that the short-term behaviors of prior policies are sharable across tasks. By evaluating the behaviors of prior policies via a hybrid value function architecture, SMEC adaptively aggregates the sharable short-term behaviors of prior policies and the long-term behaviors of the task policy, leading to coordinated decisions. Empirical results on a collection of manipulation and locomotion tasks demonstrate that SMEC outperforms existing methods, and validate the ability of SMEC to leverage related prior policies., Comment: 28 pages, 25 figures

Energy-efficient hardware implementation of machine learning algorithms for quantum computation requires nonvolatile and electrically-programmable devices, memristors, working at cryogenic temperatures that enable in-memory computing. Magnetic topological insulators are promising candidates due to their tunable magnetic order by electrical currents with high energy efficiency. Here, we utilize magnetic topological insulators as memristors (termed magnetic topological memristors) and introduce a chiral edge state-based cryogenic in-memory computing scheme. On the one hand, the chiral edge state can be tuned from left-handed to right-handed chirality through spin-momentum locked topological surface current injection. On the other hand, the chiral edge state exhibits giant and bipolar anomalous Hall resistance, which facilitates the electrical readout. The memristive switching and reading of the chiral edge state exhibit high energy efficiency, high stability, and low stochasticity. We achieve high accuracy in a proof-of-concept classification task using four magnetic topological memristors. Furthermore, our algorithm-level and circuit-level simulations of large-scale neural networks based on magnetic topological memristors demonstrate a software-level accuracy and lower energy consumption for image recognition and quantum state preparation compared with existing memristor technologies. Our results may inspire further topological quantum physics-based novel computing schemes., Comment: 33 pages, 12 figures

In transport, the topological Hall effect (THE) presents itself as non-monotonic features (or humps and dips) in the Hall signal and is widely interpreted as a sign of chiral spin textures, like magnetic skyrmions. However, when anomalous Hall effect (AHE) is also present, the co-existence of two AHEs could give rise to similar artifacts, making it difficult to distinguish between genuine THE with AHE and two-component AHE. Here we confirm genuine THE with AHE by means of transport and magneto-optical Kerr effect (MOKE) microscopy, in which magnetic skyrmions are directly observed, and find that genuine THE occurs in the transition region of the AHE. In sharp contrast, the artifact "THE", or two-component AHE occurs well beyond the saturation of the "AHE component" (under the false assumption of THE+AHE). Furthermore, we distinguish artifact "THE" from genuine THE by three methods: 1. Minor loops, 2. Temperature dependence, 3. Gate dependence. Minor loops of genuine THE with AHE are always within the full loop, while minor loops of the artifact "THE" may reveal a single loop that cannot fit into the "AHE component". Besides, the temperature or gate dependence of the artifact "THE" may also be accompanied by a polarity change of the "AHE component", as the non-monotonic features vanish, while the temperature dependence of genuine THE with AHE reveals no such change. Our work may help future researchers to exercise cautions and use these methods to examine carefully in order to ascertain genuine THE.

Symmetry breaking is a characteristic to determine which branch of a bifurcation system follows upon crossing a critical point. Specifically, in spin-orbit torque (SOT) devices, a fundamental question arises: how to break the symmetry of the perpendicular magnetic moment by the in-plane spin polarization? Here, we show that the chiral symmetry breaking by the DMI can induce the deterministic SOT switching of the perpendicular magnetization. By introducing a gradient of saturation magnetization or magnetic anisotropy, non-collinear spin textures are formed by the gradient of effective SOT strength, and thus the chiral symmetry of the SOT-induced spin textures is broken by the DMI, resulting in the deterministic magnetization switching. We introduce a strategy to induce an out-of-plane (z) gradient of magnetic properties, as a practical solution for the wafer-scale manufacture of SOT devices., Comment: 16 pages, 4 figures