Stein discrepancies have emerged as a powerful tool for retrospective improvement of Markov chain Monte Carlo output. However, the question of how to design Markov chains that are well-suited to such post-processing has yet to be addressed. This paper studies Stein importance sampling, in which weights are assigned to the states visited by a \Pi -invariant Markov chain to obtain a consistent approximation of P, the intended target. Surprisingly, the optimal choice of \Pi is not identical to the target P; we therefore propose an explicit construction for \Pi based on a novel variational argument. Explicit conditions for convergence of Stein \Pi -Importance Sampling are established. For \approx 70 % of tasks in the PosteriorDB benchmark, a significant improvement over the analogous post-processing of P -invariant Markov chains is reported.

Transformer architecture has shown impressive performance in multiple research domains and has become the backbone of many neural network models. However, there is limited understanding on how it works. In particular, with a simple predictive loss, how the representation emerges from the gradient \emph{training dynamics} remains a mystery. In this paper, for 1-layer transformer with one self-attention layer plus one decoder layer, we analyze its SGD training dynamics for the task of next token prediction in a mathematically rigorous manner. We open the black box of the dynamic process of how the self-attention layer combines input tokens, and reveal the nature of underlying inductive bias.

Breakthroughs, discoveries, and DIY tips sent every weekday. That euphoric feeling when a great idea strikes or a challenging puzzle piece fits into place is electric–and also helps our brains. Now, a team of researchers from the United States and Germany have taken a peek inside the brain to see what those so-called aha, lightbulb, or eureka moments look like. The new brain imaging shows that these flashes of insights reshape how the brain represents information and helps burn it into our memory. According to Maxi Becker, a study co-author and cognitive neuroscientist at Humboldt University in Berlin, if you have one of these aha moments when solving a problem, "you're actually more likely to remember the solution.'" The findings are detailed in a study published May 9 in the journal Nature Communications.

Common explanations for shortcut learning assume that the shortcut improves prediction only under the training distribution. Thus, models trained in the typical way by minimizing log-loss using gradient descent, which we call default-ERM, should utilize the shortcut. However, even when the stable feature determines the label in the training distribution and the shortcut does not provide any additional information, like in perception tasks, default-ERM exhibits shortcut learning. Why are such solutions preferred when the loss can be driven to zero when using the stable feature alone? By studying a linear perception task, we show that default-ERM's preference for maximizing the margin, even without overparameterization, leads to models that depend more on the shortcut than the stable feature.

Albeit revealing impressive predictive performance for several computer vision tasks, deep neural networks (DNNs) are prone to making overconfident predictions. This limits the adoption and wider utilization of DNNs in many safety-critical applications. There have been recent efforts toward calibrating DNNs, however, almost all of them focus on the classification task. Surprisingly, very little attention has been devoted to calibrating modern DNN-based object detectors, especially detection transformers, which have recently demonstrated promising detection performance and are influential in many decision-making systems. In this work, we address the problem by proposing a mechanism for calibrated detection transformers (Cal-DETR), particularly for Deformable-DETR, UP-DETR, and DINO.

A nonprofit parents coalition is calling on multiple congressional committees to launch an investigation into Meta for prioritizing engagement metrics that put children's safety at risk. The call is part of a three-pronged attack campaign by the American Parents Coalition (APC), launched Thursday. It includes a letter to lawmakers with calls for investigations, a new parental notification system to help parents stay informed on issues impacting their kids at Meta and beyond, and mobile billboards at Meta D.C. and California headquarters, calling out the company for failure to adequately prioritize protecting children. APC's campaign follows an April Wall Street Journal report that included an investigation looking into how the company's metrics focus has led to potential harms for children. "This is not the first time Meta has been caught making tech available to kids that exposes them to inappropriate content," APC Executive Director Alleigh Marre said. "Parents across America should be extremely wary of their children's online activity, especially when it involves emerging technology like AI digital companions.

FIRST ON FOX: Dozens of drones that traipsed over Langley Air Force base in late 2023 revealed an astonishing oversight: Military officials did not believe they had the authority to shoot down the unmanned vehicles over the U.S. homeland. A new bipartisan bill, known as the COUNTER Act, seeks to rectify that, offering more bases the opportunity to become a "covered facility," or one that has the authority to shoot down drones that encroach on their airspace. The new bill has broad bipartisan and bicameral support, giving it a greater chance of becoming law. It's led by Armed Services Committee members Tom Cotton, R-Ark., and Kirsten Gillibrand, D-N.Y., in the Senate, and companion legislation is being introduced by August Pfluger, R-Texas, and Chrissy Houlahan, D-Pa., in the House. Currently, only half of the 360 domestic U.S. bases are considered "covered facilities" that are allowed to engage with unidentified drones.

Image restoration involves recovering a high-quality clean image from its degraded version. Deep learning-based methods have significantly improved image restoration performance, however, they have limited generalization ability to different degradation types and levels. This restricts their real-world application since it requires training individual models for each specific degradation and knowing the input degradation type to apply the relevant model. We present a prompt-based learning approach, PromptIR, for All-In-One image restoration that can effectively restore images from various types and levels of degradation. In particular, our method uses prompts to encode degradation-specific information, which is then used to dynamically guide the restoration network.

The reliance of text classifiers on spurious correlations can lead to poor generalization at deployment, raising concerns about their use in safety-critical domains such as healthcare. In this work, we propose to use counterfactual data augmentation, guided by knowledge of the causal structure of the data, to simulate interventions on spurious features and to learn more robust text classifiers. We show that this strategy is appropriate in prediction problems where the label is spuriously correlated with an attribute. Under the assumptions of such problems, we discuss the favorable sample complexity of counterfactual data augmentation, compared to importance re-weighting. Pragmatically, we match examples using auxiliary data, based on diff-in-diff methodology, and use a large language model (LLM) to represent a conditional probability of text.

The stunning qualitative improvement of text-to-image models has led to their widespread attention and adoption. However, we lack a comprehensive quantitative understanding of their capabilities and risks. To fill this gap, we introduce a new benchmark, Holistic Evaluation of Text-to-Image Models (HEIM). Whereas previous evaluations focus mostly on image-text alignment and image quality, we identify 12 aspects, including text-image alignment, image quality, aesthetics, originality, reasoning, knowledge, bias, toxicity, fairness, robustness, multilinguality, and efficiency. We curate 62 scenarios encompassing these aspects and evaluate 26 state-of-the-art text-to-image models on this benchmark.