Jonathan M. Gitlin / Ars Technica: A study of Waymo robotaxi data for 13.8M trips from August 2023 to December 2025: the share of miles driven with a passenger onboard climbed from 36% to 56%  —  The age of robotaxis, long the preserve of science fiction, is now a reality, at least in a handful of American cities.

James Paton / Financial Times: How AI tools from Harvey, Legora, and Anthropic are challenging legal tech incumbents like Thomson Reuters and LexisNexis, which are upgrading their products  —  Incumbents are upgrading products, while emphasising authoritative content and expertise that legal teams can trust

Paige Smith / Bloomberg: Sources: OnePay, a fintech app backed by Walmart and valued at $4B+, has 6M MAUs and $50B in annualized payments, double the figures from the previous year  —  Backed by the retailer and Ribbit Capital, OnePay is reaching beyond its initial customer base to take on banks.

Krishn Kaushik / Financial Times: Taiwan and South Korea's stock market capitalizations overtake India's as global investors favor markets with companies positioned to benefit from the AI boom  —  Exchanges in Taiwan and South Korea have overtaken India's in the past week as the two countries' chipmakers surge

SpaceX set its IPO price at $135, putting the company on track for a record public listing and a valuation near $1.77 trillion. The post SpaceX Sets $135 IPO Price at $1.77T Valuation appeared first on TechRepublic.

Bloomberg: Current and former employees detail Amazon's e-commerce struggles in India, where local rivals have challenged the company more than anywhere outside of China  —  More than a decade after entering the country, the company is battling Walmart's Flipkart and other incumbents.

Financial Times: World's highest-grossing law firm Kirkland & Ellis agrees to a multiyear deal with Palantir to develop an AI tool to help advise PE firms on raising capital  —  Kirkland & Ellis has agreed a multiyear deal with Palantir to develop AI technology to help it advise private equity groups …

A screen grab captured from a Russian Defense Ministry video shows crews of Su-34 fighter-bombers preparing a FAB-500 to attack Ukraine in 2024. Russian Defense Ministry / Handout/Anadolu via Getty Images Russia's air force keeps bombing its own territories amid its high-tempo glide bomb strike campaigns. One tally found that Russia has bombed itself at least 330 times in the last three years. The friendly-fire rate shows a combination of poor procedures and fatigue, the UK's MOD said. Russia has been plagued for years by friendly-fire incidents with its glide bombs, showing flaws in how Moscow is waging its airstrike campaign, said the UK defense ministry. "These instances demonstrate continued errors in Russia's ability to successfully employ their munitions on intended targets," the ministry said in an intelligence update published on Wednesday. The update cited an independent Russian open-source intelligence group, Astra, which has for years reported on and posted footage of destruction caused by glide bombs dropped on Russian soil and occupied Ukrainian territories. The group said in late May that the Russian air force had dropped at least 25 glide bombs on its controlled territories so far this year, after 143 incidents in 2025 and another 165 in 2024. Russia began hammering Ukraine with dozens of air-launched glide bomb attacks a day in 2023, and has gradually intensified the tempo of its strikes to now drop well over 100 munitions daily. The UK defense ministry estimated that the Kremlin now conducts over 200 fighter sorties a day, delivering about 180 to 250 glide bombs daily. The repeated friendly-fire incidents are likely due to a combination of poor procedures for arming the fighters and execution mistakes by air crews, the ministry wrote. "The frequency of these occurrences likely demonstrates a degree of air and ground crew fatigue within the Russian force, as well as exposing inadequate military training," it added. Russia's defense ministry did not respond to a request for comment sent outside regular business hours by Business Insider. Russian fighter aircraft often launch glide bombs at Ukraine from inside friendly territory, using FAB general-purpose bombs modified with UPMK guidance kits. A FAB-500 with a UMPK kit mounted under a Su-34. Russian Defense Ministry/Handout/Anadolu via Getty Images Analysts said that year that the rate of friendly-fire incidents, which often injure or kill civilians in Russian border cities such as Belgorod, may have been due to the guidance kits being faulty or of poor quality. Russian officials in affected areas have acknowledged such errors before. One of the Russian defense ministry's most explicit admissions came in April 2023, when it said a Su-34 had mistakenly dropped a bomb in an "abnormal descent of aviation ammunition" on Belgorod. Local authorities said the munition left a 60-foot-wide crater and injured two women. Read the original article on Business Insider

Disney board member Bob Iger, ESPN chairman Jimmy Pitaro, and other media executives were spotted at Game 1 of the NBA Finals in San Antonio, Texas. Screengrab by Steven Tweedie. Disney's Bob Iger and Josh D'Amaro were among the high-profile attendees at Game 1 of the NBA Finals on Wednesday. The commentators for the night joked that the men made up a "murderers' row of executives." The New York Knicks defeated the San Antonio Spurs 105-95. A squad of media heavyweights turned up at the NBA Finals on Wednesday evening. Among them were Disney board member Bob Iger, Disney CEO Josh D'Amaro, ESPN chairman Jimmy Pitaro, and NBA Commissioner Adam Silver. The group was spotted at Frost Bank Center in San Antonio as the New York Knicks took on the San Antonio Spurs in the first game of the NBA Finals. Their presence later earned a shoutout on the ABC broadcast. The commentators for the night — Mike Breen, Richard Jefferson, and Tim Legler — joked that the four men made up a "murderers' row of executives" and looked extremely "handsome." They added that they weren't trying to "curry favor because they're our bosses." The NBA Finals are produced by ESPN and broadcast exclusively on ABC, both of which are owned by The Walt Disney Company. The Knicks' first trip to the NBA Finals in more than two decades brought a star-studded crowd to San Antonio, including celebrity fans Ben Stiller and Spike Lee. The night ended on a high note for Knicks supporters, as the team defeated the Spurs 105-95. Game 2 is scheduled to take place on Friday. Read the original article on Business Insider

Discussion | Link

Discussion | Link

Despite broadly connected digital infrastructure, standard fare TTPs are enough to cause trouble for Afghanistan's porous cybersecurity.

arXiv:2606.04032v1 Announce Type: new Abstract: Transformers have become the standard solution for various AI tasks, with the query, key, and value (QKV) attention formulation playing a central role. However, the individual contribution of these three projections and the impact of omitting some remain poorly understood. We systematically evaluate three projection sharing constraints: a) Q-K=V (shared key-value), b) Q=K-V (shared query-key), and c) Q=K=V (single projection). The last two variants produce symmetric attention maps; to address this, we also explore asymmetric attention via 2D positional encodings. Through experiments spanning synthetic tasks, vision (MNIST, CIFAR, TinyImageNet, anomaly), and language modeling (300M and 1.2B parameter models on 10B tokens), we discovered that our transformers perform on par or occasionally better than the QKV transformer. In language modeling, Q-K=V projection sharing achieves 50% KV cache reduction with only 3.1% perplexity degradation. Crucially, projection sharing is complementary to head sharing (GQA/MQA): combining Q-K=V with GQA-4 yields 87.5% cache reduction, while Q-K=V + MQA achieves 96.9%, enabling practical on-device inference. We show that Q-K=V preserves quality because keys and values can occupy similar representational spaces and attention operates in a low-rank regime, whereas Q=K-V breaks attention directionality. Our results systematically characterize projection sharing as an underexplored instance of weight tying in attention, with direct, quantifiable inference memory benefits, particularly valuable for edge deployment. The code is publicly available at https://github.com/anushamadan02/Do-Transformers-Need-3-Projections

arXiv:2606.04045v1 Announce Type: new Abstract: Representation learning is often described as preserving the information in an input that is relevant for prediction. This work asks what relevance means for a fixed supervised decision problem. A representation is defined to be Bayes-sufficient for a joint distribution and loss if some prediction head can use it to implement a Bayes-optimal action rule. This makes the target information loss-dependent. In the almost-surely unique Bayes-action case, the relevant object is a Bayes quotient, which identifies inputs that require the same Bayes-optimal action. A representation is sufficient when it refines this quotient, and Bayes-minimal when it is informationally equivalent to it. The framework connects naturally to property elicitation: zero-one loss requires the Bayes class, squared loss the conditional mean, Brier loss the conditional probability in binary prediction, and log loss or strictly proper scoring rules the predictive distribution. Controlled finite experiments, learned neural bottleneck experiments, and a real-data iNaturalist taxonomic refinement experiment illustrate the distinction between sufficiency, minimality, and retained non-required information. For a fixed supervised problem, the distribution and the loss determine the Bayes action, the Bayes action determines the quotient, and the quotient determines the minimal information required for Bayes-optimal prediction.

arXiv:2606.04033v1 Announce Type: new Abstract: The validation of advanced nuclear reactor designs and fuel concepts requires critical experiments with high neutronic similarity to the target technology. Neutronic similarity is quantified by the correlation coefficient $c_k$, which captures the shared bias in $k_\text{eff}$ induced by uncertainties in nuclear data. Generally, a $c_k\geq0.9$ is needed for an experiment to be sufficiently similar to a target technology. This work presents a methodology for the inverse design of critical experiments. Deep neural network surrogate modeling and nonparametric gradient optimization are used to generate experiment geometries that maximize $c_k$. A deep neural network is trained on OpenMC-calculated sensitivity vectors for grid-based critical experiment geometries. The model architecture combines a U-Net convolutional encoder-decoder with a novel multigroup attention pooling layer, introduced to capture the differing spatial dependencies of sensitivities. Multigroup attention pooling is shown to achieve better performance than traditional pooling, as well as interpretable internal behavior. The differentiability of the surrogate enables gradient-based optimization of the full combinatorial design space, allowing $c_k$ to be maximized by directly changing the material assignment of each position in the geometry grid. The method is applied to the validation of the TN-Americas TN-LC transportation cask with HALEU fuel, for which existing critical experiment coverage is limited. The optimization procedure is shown to produce experiment geometries achieving $c_k$ scores of 0.97757, 0.81324, and 0.93276 for three configurations of interest. This approach demonstrates the potential of deep learning and gradient optimization to accelerate the development of advanced nuclear technology.

arXiv:2606.04107v1 Announce Type: new Abstract: Single-image reflection separation is highly challenging under extreme conditions like glare or weak reflections. Existing methods often struggle to recover both layers in glare or weak-reflection scenarios because of insufficient information. This paper presents a diffusion model explicitly fine-tuned for this task, leveraging generative diffusion priors for robust separation. Our method simultaneously generates transmission and reflection layers through a unified diffusion model, incorporating a novel cross-layer self-attention mechanism for better feature disentanglement. We further introduce a disjoint sampling strategy to iteratively reduce interference between the layers during diffusion and a latent optimization step with a learned composition function for improved results in complex real-world scenarios. Extensive experiments demonstrate that our approach surpasses state-of-the-art methods on multiple real-world benchmarks. Project page: https://brian90709.github.io/diff-reflection-separation/

arXiv:2606.04061v1 Announce Type: new Abstract: Large-scale web-harvested datasets have fueled the progress of cross-modal retrieval but inevitably suffer from noisy correspondence, which severely degrades model generalization. Existing methods primarily address this by filtering out noise or seeking a substitute label, yet they predominantly remain bound by a "Discrete Selection" paradigm. We argue that relying on a single discrete proxy induces Single-Point Fragility and Discretization Error. To overcome these limitations, we propose a novel framework, Intra-modal Neighbor-aware Noise Rectification (IN2R), which shifts the paradigm from searching for a substitute to synthesizing a reliable supervision target. Leveraging the intrinsic geometric stability of intra-modal data, IN2R employs a Graph Refiner to perform relational reasoning over neighbors retrieved from a dynamic Cross-Model Memory. Instead of propagating discrete labels, our method synthesizes a continuous, soft prototype that reflects the consensus of the local semantic neighborhood, effectively rectifying inter-modal misalignment. Extensive experiments on Flickr30K, MS-COCO, and CC152K demonstrate that IN2R significantly outperforms state-of-the-art methods. Our code and pre-trained models are publicly available at https://github.com/liuyyy111/IN2R.

arXiv:2606.04189v1 Announce Type: new Abstract: Aspect-Based Sentiment Analysis (ABSA) requires high-quality datasets to train reliable models. However, existing annotation tools treat output as flat files, leaving researchers to manually consolidate multi-annotator data, reconstruct relational structures, and compute reliability metrics through custom scripts. This paper introduces ACAT (Aspect-based sentiment analysis Collaborative Annotation Tool), a web-based platform natively supporting four ABSA workflows: (1) Aspect-Category Sentiment Analysis, (2) Clause-Level Segmentation, (3) Aspect-Term Sentiment Analysis with character-level position tracking, and (4) Aspect Sentiment Triplet Extraction with dual span offset preservation. Its core contribution is an automated Extract, Transform, Load (ETL) pipeline that aligns collaborative annotations and computes Inter-Annotator Agreement (IAA) metrics directly at export, yielding training-ready datasets. In a preliminary validation on 1,002 restaurant reviews with two annotators of differing expertise, ACAT achieves a median annotation time of 31.58 seconds and a raw IAA ranging from 0.78 to 0.86 across all tasks.

arXiv:2606.04029v1 Announce Type: new Abstract: Reinforcement Learning (RL) has received increasing attention and adoption in real-world use cases. Most of these systems follow a train-then-fix paradigm, where trained agents do not learn while interacting with the world until performance degrades and retraining becomes necessary. In this position paper, we argue that deploying an agent that is incapable of optimality, but receives an evaluative reward signal, is inherently a continual RL problem. We identify four sources of non-stationarity after deployment that necessitate never-ending learning, and highlight why the best deployed agents never stop adapting. We analyze successful examples of continual RL in the real world, and present the community with the advantages and measures to move away from the current train-then-fix paradigm.

arXiv:2606.04031v1 Announce Type: new Abstract: Coupled gradient descent--where the update of one parameter block depends on another--underlies bilevel optimization, two-time-scale stochastic approximation, and adversarial training. When the coupled Jacobian is block-triangular, asymptotic stability is governed by the spectral radii of the diagonal blocks, yet transient amplification before convergence can be arbitrarily large due to non-normality. We develop a sharp pseudospectral theory for such block-triangular Jacobians, proving that the Kreiss constant satisfies $K(J) \leq 2/(1-\gamma) + \|C\|/(4(1-\gamma))$ when the diagonal blocks are symmetric with spectral radii at most $\gamma < 1$, and we establish matching minimax lower bounds. We characterize the critical coupling threshold for spectral instability and extend the analysis to nearly self-referential systems via a Neumann-series perturbation framework. As a consequence, we obtain a finite-horizon iteration-complexity bound of $O(K(J)^2 \log(1/\delta))$ for stochastic coupled descent. Framed as scaling laws for non-stationary two-time-scale optimization, our results expose a non-asymptotic, instance-dependent regime of high-dimensional learning dynamics that is invisible to spectral-radius analysis. Experiments on linear-quadratic problems, IQC-based comparisons, and neural-network training confirm the theory.