Search for: All records

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/. 

Hypervalent iodine compounds are a widely used class of metal-free oxidants that find application in organic synthesis. Due to the homology between the reactivity of hypervalent iodine and many transition metals ¾ oxidative addition, ligand exchange, and reductive elimination can be facile for both ¾ hypervalent iodine species find application in a variety of synthetically important organic transformations. Major limitations of these reagents include the frequent need for (super)stoichiometric loading and the intrinsically poor atom economy that results from the generation of stoichiometric quantities of iodoarene byproducts. In addition, hypervalent iodine reagents are often synthesized using metal-based terminal oxidants, which compound the resulting waste stream. Recently, substantial progress has been made to address these limitations. Here, we discuss progress towards sustainable synthetic methods for the preparation of hypervalent iodine compounds and application of those methods in the context of hypervalent iodine catalysis. The discussion is organized according to the active oxygen content, and thus atom economy, of the terminal oxidant employed. Hypervalent iodine electrochemistry and the development of recyclable iodoarenes are also discussed. 

Diatoms are a group of single-celled photosynthetic algae that use biochemical pathways to bio-mineralize and self-assemble three-dimensional photonic crystals with unique photonic and micro- & nano-fluidic properties. In recent years, diatom biosilica has been used in surface-enhanced Raman scattering (SERS) based optofluidic sensors for detection of a variety of chemical and biological molecules. In this paper, we present a study to develop a microfluidic pumping system using super-hydrophilic diatom thin films. The desire to develop such a system stems from the requirement to create a low-cost, self-powered microfluidic pumping system that can sustain a continuous flow over an extended period of time. The diatom biosilica acts not only as the driving force behind the flow, but also serves as ultra-sensitive SERS substrates that allows for trace detection of various molecules. Liquid is drawn from a reservoir to the tip of a 150μm inner diameter capillary tube positioned directly over the diatom film. A thin and long horizontal reservoir is used to prevent flooding on the diatom film when the liquid is initially drawn to the diatom film through a capillary tube from the reservoir. The connection of the meniscus from the capillary to the film was maintained from a horizontal reservoir for a recorded time of 20 hours and 32 minutes before the partially filled reservoir emptied. Flow rates of 0.38, 0.22 and 0.16µL/min were achieved for square biosilica thin films of 49mm2, 25mm2, and 9mm2 at a temperature of 63̊F and 45% relative humidity respectively. A temperature-controlled system was introduced for the 49mm2 substrate and flow rates of 0.60, 0.82, 0.93, and 1.15µL/min were observed at 72, 77, 86, and 95̊F at 21% relative humidity respectively. More testing and analysis will be performed to test the operation limits of the proposed self-powered microfluidic system.