Neural attention has become central to many state-of-the-art models in natural language processing and related domains. Attention networks are an easy-to-train and effective method for softly simulating alignment; however, the approach does not marginalize over latent alignments in a probabilistic sense. This property makes it difficult to compare attention to other alignment approaches, to compose it with probabilistic models, and to perform posterior inference conditioned on observed data. A related latent approach, hard attention, fixes these issues, but is generally harder to train and less accurate. This work considers variational attention networks, alternatives to soft and hard attention for learning latent variable alignment models, with tighter approximation bounds based on amortized variational inference. We further propose methods for reducing the variance of gradients to make these approaches computationally feasible. Experiments show that for machine translation and visual question answering, inefficient exact latent variable models outperform standard neural attention, but these gains go away when using hard attention based training. On the other hand, variational attention retains most of the performance gain but with training speed comparable to neural attention.

Graves being prepared for the victims of an airstrike on a school in Minab in southern Iran, 2 March 2026. Graves being prepared for the victims of an airstrike on a school in Minab in southern Iran, 2 March 2026. A photo of Iran's bombed schoolgirl graveyard went around the world. T he graves, freshly dug, lie in neat rows of 20 across. More than 60 have already been carved out of the earth, with a few clusters of people standing gathered around them.

In 2025, when Elon Musk joined the government as the de facto head of something called the "department of government efficiency", he declared that governments were poorly configured "big dumb machines". To the senator Ted Cruz, he explained that "the only way to reconcile the databases and get rid of waste and fraud is to actually look at the computers". Muskism came to Washington soaked in memes, adolescent boasts and sadistic victory dances over mass firings. Leading a team of teenage coders and mid-level managers drawn from his suite of companies, Musk aimed to enter the codebase and rewrite regulations and budget lines from within. He would drag the paper-pushing bureaucracy kicking and screaming into the digital 21st century, scanning the contents of cavernous rooms of filing cabinets and feeding the data into a single interoperable system. The undertaking combined features of private equity-led restructuring with startup management, shot through with the sensibility of gaming and rightwing culture war. To succeed, he would need "God mode", an overview of the whole. If the mandate of Doge was to "[modernise] federal technology and software to maximise governmental efficiency and productivity", in the words of the executive order that launched the initiative on 20 January 2025, the reality was a strengthening of the state's surveillance capacities. Over time, Musk had become convinced that the real bugs in the code were people, especially the non-white illegal immigrants whom he saw as pawns in a liberal scheme to corrupt democracy and beneficiaries of what he called "suicidal empathy". He understood empathy itself in coding terms.

Recent work used importance sampling ideas for better variational bounds on likelihoods. We clarify the applicability of these ideas to pure probabilistic inference, by showing the resulting Importance Weighted Variational Inference (IWVI) technique is an instance of augmented variational inference, thus identifying the looseness in previous work. Experiments confirm IWVI's practicality for probabilistic inference. As a second contribution, we investigate inference with elliptical distributions, which improves accuracy in low dimensions, and convergence in high dimensions.

Tennessee minors sue Musk's xAI, alleging Grok generated sexual images of them Governments and regulators around the world have launched probes into xAI, imposed bans and demanded safeguards in a growing push to curb illegal and offensive material. Three Tennessee plaintiffs, including two minors, sued Elon Musk's xAI on Monday, alleging that it knowingly designed its Grok image generator to let people create sexually explicit content by using real photos of others. The lawsuit, filed in the San Jose, California federal court, is seeking class-action status for people in the United States who were reasonably identifiable in sexualized images or videos generated by Grok based on real images of themselves. The artificial intelligence company did not immediately respond to a request for comment. After an outcry over sexually explicit content generated by the chatbot, xAI said in January that it had blocked all users from editing images of real people in revealing clothing and from generating images of people in revealing clothing in jurisdictions where it's illegal. Governments and regulators around the world have also since launched probes, imposed bans and demanded safeguards in a growing push to curb illegal and offensive material.

In this paper, we use a linear program to empirically decompose fitted neural networks into ensembles of low-bias sub-networks. We show that these sub-networks are relatively uncorrelated which leads to an internal regularization process, very much like a random forest, which can explain why a neural network is surprisingly resistant to overfitting. We then demonstrate this in practice by applying large neural networks, with hundreds of parameters per training observation, to a collection of 116 real-world data sets from the UCI Machine Learning Repository. This collection of data sets contains a much smaller number of training examples than the types of image classification tasks generally studied in the deep learning literature, as well as non-trivial label noise. We show that even in this setting deep neural nets are capable of achieving superior classification accuracy without overfitting.

The Equalized Odds (for short, EO) is one of the most popular measures of discrimination used in the supervised learning setting. It ascertains fairness through the balance of the misclassification rates (false positive and negative) across the protected groups -- e.g., in the context of law enforcement, an African-American defendant who would not commit a future crime will have an equal opportunity of being released, compared to a non-recidivating Caucasian defendant. Despite this noble goal, it has been acknowledged in the literature that statistical tests based on the EO are oblivious to the underlying causal mechanisms that generated the disparity in the first place (Hardt et al. 2016). This leads to a critical disconnect between statistical measures readable from the data and the meaning of discrimination in the legal system, where compelling evidence that the observed disparity is tied to a specific causal process deemed unfair by society is required to characterize discrimination. The goal of this paper is to develop a principled approach to connect the statistical disparities characterized by the EO and the underlying, elusive, and frequently unobserved, causal mechanisms that generated such inequality. We start by introducing a new family of counterfactual measures that allows one to explain the misclassification disparities in terms of the underlying mechanisms in an arbitrary, non-parametric structural causal model. This will, in turn, allow legal and data analysts to interpret currently deployed classifiers through causal lens, linking the statistical disparities found in the data to the corresponding causal processes. Leveraging the new family of counterfactual measures, we develop a learning procedure to construct a classifier that is statistically efficient, interpretable, and compatible with the basic human intuition of fairness. We demonstrate our results through experiments in both real (COMPAS) and synthetic datasets.

This paper presents a new class of gradient methods for distributed machine learning that adaptively skip the gradient calculations to learn with reduced communication and computation. Simple rules are designed to detect slowly-varying gradients and, therefore, trigger the reuse of outdated gradients. The resultant gradient-based algorithms are termed Lazily Aggregated Gradient --- justifying our acronym LAG used henceforth. Theoretically, the merits of this contribution are: i) the convergence rate is the same as batch gradient descent in strongly-convex, convex, and nonconvex cases; and, ii) if the distributed datasets are heterogeneous (quantified by certain measurable constants), the communication rounds needed to achieve a targeted accuracy are reduced thanks to the adaptive reuse of lagged gradients. Numerical experiments on both synthetic and real data corroborate a significant communication reduction compared to alternatives.

Data-driven approaches to solving robotic tasks have gained a lot of traction in recent years. However, most existing policies are trained on large-scale datasets collected in curated lab settings. If we aim to deploy these models in unstructured visual environments like people's homes, they will be unable to cope with the mismatch in data distribution. In such light, we present the first systematic effort in collecting a large dataset for robotic grasping in homes. First, to scale and parallelize data collection, we built a low cost mobile manipulator assembled for under 3K USD.