Search for: All records

Proceedings of The 28th International Conference on Artificial Intelligence and Statistics, Proceedings of Machine Learning Research #258 

We introduce RandAR, a decoder-only visual autoregressive (AR) model capable of generatng images in arbitrary token orders. Unlike previous decoder-only AR models that rely on a predefined generation order, RandAR removes this inductive bias, unlocking new capabilities in decoder-only generation. Our essential design enabling random order is to insert a "position instruction token" before each image token to be predicted, representing the spatial location of the next image token. Trained on randomly permuted token sequences -- a more challenging task than fixed-order generation, RandAR achieves comparable performance to conventional raster-order counterpart. More importantly, decoder-only transformers trained from random orders acquire new capabilities. For the efficiency bottleneck of AR models, RandAR adopts parallel decoding with KV-Cache at inference time, enjoying 2.5x acceleration without sacrificing generation quality. Additionally, RandAR supports in-painting, outpainting and resolution extrapolation in a zero-shot manner.We hope RandAR inspires new directions for decoder-only visual generation models and broadens their applications across diverse scenarios. Our project page is at https://rand-ar.github.io/. 

Abstract In the present work we revisit the problem of the quantum droplet in atomic Bose–Einstein condensates with an eye towards describing its ground state in the large density, so-called Thomas–Fermi (TF) limit. We consider the problem as being separable into 3 distinct regions: an inner one, where the TF approximation is valid, a sharp transition region where the density abruptly drops towards the (vanishing) background value and an outer region which asymptotes to the background value. We analyze the spatial extent of each of these regions, and develop a systematic effective description of the rapid intermediate transition region. Accordingly, we derive a uniformly valid description of the ground state that is found to accurately match our numerical computations. As an additional application of our considerations, we show that this formulation allows for an analytical approximation of excited states such as the (trapped) dark soliton in the large density limit. 

We consider a discrete-time system where a resource-constrained source (e.g., a small sensor) transmits its time-sensitive data to a destination over a time-varying wireless channel. Each transmission incurs a fixed transmission cost (e.g., energy cost), and no transmission results in a staleness cost represented by the Age-of-Information. The source must balance the tradeoff between transmission and staleness costs. To address this challenge, we develop a robust online algorithm to minimize the sum of transmission and staleness costs, ensuring a worst-case performance guarantee. While online algorithms are robust, they are usually overly conservative and may have a poor average performance in typical scenarios. In contrast, by leveraging historical data and prediction models, machine learning (ML) algorithms perform well in average cases. However, they typically lack worst-case performance guarantees. To achieve the best of both worlds, we design a learning-augmented online algorithm that exhibits two desired properties: (i) consistency: closely approximating the optimal offline algorithm when the ML prediction is accurate and trusted; (ii) robustness: ensuring worst case performance guarantee even ML predictions are inaccurate. Finally, we perform extensive simulations to show that our online algorithm performs well empirically and that our learning augmented algorithm achieves both consistency and robustness. 

Abstract Because of the extreme purity, lack of disorder, and complex order parameter, the first-order superfluid 3 He A–B transition is the leading model system for first order transitions in the early universe. Here we report on the path dependence of the supercooling of the A phase over a wide range of pressures below 29.3 bar at nearly zero magnetic field. The A phase can be cooled significantly below the thermodynamic A–B transition temperature. While the extent of supercooling is highly reproducible, it depends strongly upon the cooling trajectory: The metastability of the A phase is enhanced by transiting through regions where the A phase is more stable. We provide evidence that some of the additional supercooling is due to the elimination of B phase nucleation precursors formed upon passage through the superfluid transition. A greater understanding of the physics is essential before 3 He can be exploited to model transitions in the early universe.