First and second authors contributed equally to this work Abstract --Modern recommenders describe each item with hundreds of sparse semantic tags, yet most quantum pipelines still map one qubit per tag, demanding well beyond one hundred qubits, far out of reach for current noisy-intermediate-scale quantum (NISQ) devices and prone to deep, error-amplifying circuits. We close this gap with a three-stage hybrid machine learning algorithm that compresses tag profiles, optimizes feature selection under a fixed qubit budget via QAOA, and scores recommendations with a Quantum semi-Random Forest (QsRF) built on just five qubits, while performing similarly to the state-of-the-art methods. Leveraging SVD sketching and k-means, we learn a 1 000-atom dictionary ( >97 % variance), then solve a 20 20 QUBO via depth-3 QAOA to select 5 atoms. A 100-tree QsRF trained on these codes matches full-feature baselines on ICM-150/500. To compress this combinatorial explosion, recent hybrid pipelines formulate feature selection as a Q uadratic U nconstrained Binary O ptimisation (QUBO) and delegate the search to quantum annealers [1], [2] or shallow gate-based circuits [3].

Tran* sat across from me, phone in hand, scrolling. "I just wanted to make sure I didn't say the wrong thing," he explained, referring to a disagreement with his partner. "So I asked ChatGPT what I should say." He read the chatbot-generated message aloud. It was articulate, logical and composed – too composed.

In 2023, a Belgian man reportedly ended his life after developing eco-anxiety and confiding in an AI chatbot over six weeks about the future of the planet. Without those conversations, his widow reportedly told the Belgian outlet La Libre, "he would still be here". In April this year, a 35-year-old Florida man was shot and killed by police in another chatbot-related incident: his father later told media that the man had come to believe an entity named Juliet was trapped inside ChatGPT, and then killed by OpenAI. When the man, who reportedly struggled with bipolar disorder and schizophrenia, was confronted by police, he allegedly charged at them with a knife. The wide availability of chatbots in the past few years has apparently led some to believe there is a ghost in the machine – one that is conscious, capable of loving and being loved.

When the air ambulance brought Jim Ashworth-Beaumont to King's College hospital in south-east London, nobody thought he had a hope. He had been cycling home when a lorry driver failed to spot him alongside his trailer while turning left after a set of traffic lights. The vehicle's wheels opened his torso like a sardine tin, puncturing his lungs and splitting his liver in two. They also tore off his right arm. Weeks after the accident, in July 2020, Ashworth-Beaumont would see a photo of the severed limb taken by a doctor while it lay beside him in hospital. He had asked to see the picture and says it helped him come to terms with his loss. "My hand didn't look too bad," he says. "It was as if it was waving goodbye to me." Ashworth-Beaumont, a super-fit and sunny former Royal Marine from Edinburgh, would go on to spend six weeks in an induced coma as surgeons raced to repair his crushed body. But as he lay on the road, waiting for the paramedics, his only thoughts were that he was dying.

Breakthroughs, discoveries, and DIY tips sent every weekday. On July Fourth, amid a cacophony of fireworks and flame-throwing propane grills, a seemingly ordinary lightning strike hit somewhere in Grand Canyon National Park. The resulting spark ignited surrounding dry vegetation, and strong winds quickly spread the flames for miles. Over the course of several weeks, that initial spark has grown into a blaze engulfing more than 100,000 acres, officially classifying it as a "megafire" and the largest wildfire of 2025…so far. As of this writing, "The Dragon Bravo Fire" has already destroyed 70 buildings, including the historic Grand Canyon Lodge.

Recent advances in large language models have demonstrated considerable potential in scientific domains such as drug repositioning. However, their effectiveness remains constrained when reasoning extends beyond the knowledge acquired during pre-training. Conventional approaches, such as fine-tuning or retrieval-augmented generation, face limitations in either imposing high computational overhead or failing to fully exploit structured scientific data. To overcome these challenges, we propose DrugM-CTS, a novel framework that synergistically integrates RAG, multi-agent collaboration, and Monte Carlo Tree Search for drug repositioning. The framework employs five specialized agents tasked with retrieving and analyzing molecular and protein information, thereby enabling structured and iterative reasoning. Extensive experiments on the DrugBank and KIBA datasets demonstrate that DrugMCTS achieves substantially higher recall and robustness compared to both general-purpose LLMs and deep learning baselines. Our results highlight the importance of structured reasoning, agent-based collaboration, and feedback-driven search mechanisms in advancing LLM applications for drug repositioning.

Human action recognition plays a critical role in healthcare and medicine, supporting applications such as patient behavior monitoring, fall detection, surgical robot supervision, and procedural skill assessment. While traditional models like CNNs and RNNs have achieved moderate success, they often struggle to generalize across diverse and complex actions. Recent advancements in vision-language models, especially the transformer-based CLIP model, offer promising capabilities for generalizing action recognition from video data. In this work, we evaluate CLIP on the UCF-101 dataset and systematically analyze its performance under three masking strategies: (1) percentage-based and shape-based black masking at 10%, 30%, and 50%, (2) feature-specific masking to suppress bias-inducing elements, and (3) isolation masking that retains only class-specific regions. Our results reveal that CLIP exhibits inconsistent behavior and frequent misclassifications, particularly when essential visual cues are obscured. To overcome these limitations, we propose incorporating class-specific noise, learned via a custom loss function, to reinforce attention to class-defining features. This enhancement improves classification accuracy and model confidence while reducing bias. We conclude with a discussion on the challenges of applying such models in clinical domains and outline directions for future work to improve generalizability across domain-independent healthcare scenarios.

Adm. Daryl Caudle joins'America's Newsroom' to discuss rising tensions with China's navy, the use of AI in US defense, and a powerful Memorial Day re-enlistment ceremony at Ground Zero. The U.S. Navy has called off a search for a sailor assigned to the USS George Washington amid reports that he possibly went overboard while the ship was sailing north of Australia. The sailor was reported overboard on the aircraft carrier on Monday as the ship was transiting the Timor Sea, the Navy said. US DEFENSE OFFICIAL REACTS TO IRAN'S CLAIMS ABOUT ENCOUNTER WITH WARSHIP This photo shows a general view of U.S. aircraft carrier USS George Washington shortly after berthing at Manila Bay in Manila on July 3. (TED ALJIBE/AFP via Getty Images) The search effort involving the George Washington, its carrier strike group, as well as the Australian Defence (sic) Force and Australian Border Force, concluded at 12:40 p.m. Wednesday. "USS George Washington expresses sincere condolences to those impacted by this loss and is actively engaged with the crew to make services available to tend to their needs during this challenging time," Lt. Cmdr.

We present MOSS, a multi-objective optimization framework for constructing stable sets of decision rules. MOSS incorporates three important criteria for interpretability: sparsity, accuracy, and stability, into a single multi-objective optimization framework. Importantly, MOSS allows a practitioner to rapidly evaluate the trade-off between accuracy and stability in sparse rule sets in order to select an appropriate model. We develop a specialized cutting plane algorithm in our framework to rapidly compute the Pareto frontier between these two objectives, and our algorithm scales to problem instances beyond the capabilities of commercial optimization solvers. Our experiments show that MOSS outperforms state-of-the-art rule ensembles in terms of both predictive performance and stability.

Large language models and deep learning models designed for code intelligence have revolutionized the software engineering field due to their ability to perform various code-related tasks. These models can process source code and software artifacts with high accuracy in tasks such as code completion, defect detection, and code summarization; therefore, they can potentially become an integral part of modern software engineering practices. Despite these capabilities, robustness remains a critical quality attribute for deep-code models as they may produce different results under varied and adversarial conditions (e.g., variable renaming). Metamorphic testing has become a widely used approach to evaluate models' robustness by applying semantic-preserving transformations to input programs and analyzing the stability of model outputs. While prior research has explored testing deep learning models, this systematic literature review focuses specifically on metamorphic testing for deep code models. By studying 45 primary papers, we analyze the transformations, techniques, and evaluation methods used to assess robustness. Our review summarizes the current landscape, identifying frequently evaluated models, programming tasks, datasets, target languages, and evaluation metrics, and highlights key challenges and future directions for advancing the field.