The ability to approach the same problem from different angles is a cornerstone of human intelligence that leads to robust solutions and effective adaptation to problem variations. In contrast, current RL methodologies tend to lead to policies that settle on a single solution to a given problem, making them brittle to problem variations. Replicating human flexibility in reinforcement learning agents is the challenge that we explore in this work. We tackle this challenge by extending state-of-the-art approaches to introduce DUPLEX, a method that explicitly defines a diversity objective with constraints and makes robust estimates of policies' expected behavior through successor features. The trained agents can (i) learn a diverse set of near-optimal policies in complex highly-dynamic environments and (ii) exhibit competitive and diverse skills in out-of-distribution (OOD) contexts. Empirical results indicate that DUPLEX improves over previous methods and successfully learns competitive driving styles in a hyper-realistic simulator (i.e., GranTurismo 7) as well as diverse and effective policies in several multi-context robotics MuJoCo simulations with OOD gravity forces and height limits. To the best of our knowledge, our method is the first to achieve diverse solutions in complex driving simulators and OOD robotic contexts.

The ability to approach the same problem from different angles is a cornerstone of human intelligence that leads to robust solutions and effective adaptation to problem variations. In contrast, current RL methodologies tend to lead to policies that settle on a single solution to a given problem, making them brittle to problem variations. Replicating human flexibility in reinforcement learning agents is the challenge that we explore in this work.

The ability to approach the same problem from different angles is a cornerstone of human intelligence that leads to robust solutions and effective adaptation to problem variations. In contrast, current RL methodologies tend to lead to policies that settle on a single solution to a given problem, making them brittle to problem variations. Replicating human flexibility in reinforcement learning agents is the challenge that we explore in this work. We tackle this challenge by extending state-of-the-art approaches to introduce DUPLEX, a method that explicitly defines a diversity objective with constraints and makes robust estimates of policies' expected behavior through successor features. The trained agents can (i) learn a diverse set of near-optimal policies in complex highly-dynamic environments and (ii) exhibit competitive and diverse skills in out-of-distribution (OOD) contexts.

Amazon's long-awaited update to its assistant is almost here. About 18 months after the company first previewed the "next-gen Alexa" built with generative AI, it unveiled Alexa, and early access will be available starting in March. Alexa will exist alongside the older Alexa and will cost 20 a month, unless you have a Prime membership, which will make it free to use. The new assistant will come with all the modern upgrades that its contemporaries like the redesigned Siri or Gemini offer, like more conversational interaction, better contextual understanding and the ability to "summarize complex topics" and "make suggestions based on your interests." But it does one thing differently, and it's the way Amazon purports to integrate with third-party apps and the rest of the internet that could set it apart.

Link prediction for directed graphs is a crucial task with diverse real-world applications. Recent advances in embedding methods and Graph Neural Networks (GNNs) have shown promising improvements. However, these methods often lack a thorough analysis of embedding expressiveness and suffer from ineffective benchmarks for a fair evaluation. In this paper, we propose a unified framework to assess the expressiveness of existing methods, highlighting the impact of dual embeddings and decoder design on performance. To address limitations in current experimental setups, we introduce DirLinkBench, a robust new benchmark with comprehensive coverage and standardized evaluation. The results show that current methods struggle to achieve strong performance on the new benchmark, while DiGAE outperforms others overall. We further revisit DiGAE theoretically, showing its graph convolution aligns with GCN on an undirected bipartite graph. Inspired by these insights, we propose a novel spectral directed graph auto-encoder SDGAE that achieves SOTA results on DirLinkBench. Finally, we analyze key factors influencing directed link prediction and highlight open challenges.

Large language models (LLMs) have emerged due to their capability to generate high-quality content across diverse contexts. To reduce their explosively increasing demands for computing resources, a mixture of experts (MoE) has emerged. The MoE layer enables exploiting a huge number of parameters with less computation. Applying state-of-the-art continuous batching increases throughput; however, it leads to frequent DRAM access in the MoE and attention layers. We observe that conventional computing devices have limitations when processing the MoE and attention layers, which dominate the total execution time and exhibit low arithmetic intensity (Op/B). Processing MoE layers only with devices targeting low-Op/B such as processing-in-memory (PIM) architectures is challenging due to the fluctuating Op/B in the MoE layer caused by continuous batching. To address these challenges, we propose Duplex, which comprises xPU tailored for high-Op/B and Logic-PIM to effectively perform low-Op/B operation within a single device. Duplex selects the most suitable processor based on the Op/B of each layer within LLMs. As the Op/B of the MoE layer is at least 1 and that of the attention layer has a value of 4-8 for grouped query attention, prior PIM architectures are not efficient, which place processing units inside DRAM dies and only target extremely low-Op/B (under one) operations. Based on recent trends, Logic-PIM adds more through-silicon vias (TSVs) to enable high-bandwidth communication between the DRAM die and the logic die and place powerful processing units on the logic die, which is best suited for handling low-Op/B operations ranging from few to a few dozens. To maximally utilize the xPU and Logic-PIM, we propose expert and attention co-processing.

Current directed graph embedding methods build upon undirected techniques but often inadequately capture directed edge information, leading to challenges such as: (1) Suboptimal representations for nodes with low in/out-degrees, due to the insufficient neighbor interactions; (2) Limited inductive ability for representing new nodes post-training; (3) Narrow generalizability, as training is overly coupled with specific tasks. In response, we propose DUPLEX, an inductive framework for complex embeddings of directed graphs. It (1) leverages Hermitian adjacency matrix decomposition for comprehensive neighbor integration, (2) employs a dual GAT encoder for directional neighbor modeling, and (3) features two parameter-free decoders to decouple training from particular tasks. DUPLEX outperforms state-of-the-art models, especially for nodes with sparse connectivity, and demonstrates robust inductive capability and adaptability across various tasks. The code is available at https://github.com/alipay/DUPLEX.

After years of singleminded worship of the false god Virtual Assistant, the company is rushing its AI strategy as its competitors join their hands and raise their pitchforks. The irony is it's all happening because Google thought it had the pitchfork market cornered. See, in 2017, Google researchers published the article "Attention is all you need," introducing the concept of the transformer and vastly improving the capabilities of machine learning models. You don't need to know the technical side of it (and indeed I am not the one to teach you), but it has been enormously influential and empowering; let it suffice to say that it's the T in GPT. You may well ask, why did Google give this wonderful thing away freely? While big private research outfits have been criticized in the past for withholding their work, the trend over the last few years has been toward publishing.

Google is shutting down Duplex on the Web, its AI-powered set of services that navigated sites to simplify the process of ordering food, purchasing movie tickets and more. According to a note on a Google support page, Google on the Web and any automation features enabled by it will no longer be supported as of this month. "As we continue to improve the Duplex experience, we're responding to the feedback we've heard from users and developers about how to make it even better," a Google spokesperson told TechCrunch via email, adding that Duplex on the Web partners have been notified to help them prepare for the shutdown. "By the end of this year, we'll turn down Duplex on the Web and fully focus on making AI advancements to the Duplex voice technology that helps people most every day." Google introduced Duplex on the Web, an outgrowth of its call-automating Duplex technology, during its 2019 Google I/O developer conference. To start, it was focused on a couple of narrow use cases, including opening a movie theater chain's website to fill out all of the necessary information on a user's behalf -- pausing to prompt for choices like seats.