Sooksaard/Shutterstock.com Researchers at the University of Cambridge have developed what they describe as a fundamentally new type of vaccine using artificial intelligence (AI). The vaccine’s key component was designed entirely by AI and has now been tested in people for the first time. The goal is ambitious: a single vaccine that works not just against all known human coronavirus variants, but against related bat viruses that could jump from animals to humans and cause future pandemics. Traditional vaccines train our immune system to recognise one specific virus. The problem is that viruses mutate. When they change enough, the vaccine stops working, which is why we need a new flu shot every year and why COVID vaccines have been updated repeatedly since 2021. AI offers a way around this. By analysing genetic data from thousands of related viruses, it can identify the parts that stay the same across different strains and that are unlikely to change over time. Target those stable features, and you have a vaccine that should work against the whole family, not just the strain you started with. This is exactly what the Cambridge team did. They used AI to scan viruses from the sarbecovirus family, which includes the viruses that cause both SARS and COVID, as well as a range of animal coronaviruses – looking for shared features that evolution has left largely untouched. Those features became the basis of the vaccine. DNA vaccines While many people are familiar with the mRNA shots used during the pandemic, this new vaccine uses DNA. DNA vaccines are generally more stable than mRNA vaccines, making them easier to store and transport. A significant advantage in lower-income countries where “cold-chain” infrastructure is limited. They can also be administered without needles. A high-pressure stream of liquid delivers the vaccine through the skin, making administration less painful and easier to scale up during an outbreak. DNA and RNA viruses explained. Could it protect against future pandemics? These practical advantages matter most if the vaccine itself can do something no existing jab can: protect against viruses we haven’t encountered yet. Broad-spectrum vaccines could change the way the world responds to emerging infectious diseases. By offering much wider protection than traditional vaccines, they could provide rapid immunity against new and emerging viral threats. This would equip public health officials with tools to stop future outbreaks in their tracks before they have a chance to turn into global pandemics. They could also transform our approach to more familiar diseases. Influenza is a prime target because it exists in many different strains and evolves so rapidly. Scientists have to predict which strains will dominate each flu season, and they guess wrong, vaccine effectiveness can suffer. A universal flu vaccine that targets features shared across multiple strains could eventually end the annual race to keep up with the virus. And the Ebola virus shows why this matters right now. The recent outbreak in the Democratic Republic of the Congo and Uganda is driven by the Bundibugyo strain, which bypasses existing vaccines. While researchers rush to create a new vaccine specifically for this strain, local communities remain at high risk. A broad-spectrum vaccine designed to cover an entire virus family could transform that picture. What the trial found This is the first human trial of an AI-designed vaccine. The results showed that this DNA vaccine was able to stimulate the immune system to produce antibodies that can recognise different types of sarbecoviruses. The technology was found to be safe and well tolerated. This is an exciting advance because it demonstrates how AI has the potential to design variant-proof vaccines against future pandemic threats. The needle-free delivery system could also make the vaccine easier to administer and distribute worldwide. However, there is more work to do. Although the results in this study are encouraging, the immune responses following vaccination were modest. It was also uncertain how long the protection lasts and whether further boosters will be required. Larger trials are also needed to determine whether the vaccine can prevent or reduce virus infections in the real world. A universal vaccine remains a few years away. And any new vaccine must still pass larger trials to prove it is safe, effective and provides lasting protection. But this study shows the goal is getting closer – and AI may help us get there faster. Neil Mabbott currently receives funding from the UK Biotechnology and Biological Research Council, The Creutzfeldt-Jakob Disease Foundation, and Zoetis.

Joshua Blanchard/Getty; CBS Photo Archive/Getty; Noam Galai/Getty; Tyler Le/BI On Sunday, the New York Times published an explosive interview with "60 Minutes" correspondent Scott Pelley, his first since he was fired from CBS News last week. In it, Pelley detailed a series of serious allegations — most notably, that editor in chief Bari Weiss "was putting a thumb on the scale on behalf" of the Trump administration — that dominated the news cycle that followed. But as I listened to the hourlong conversation, what I found most striking were his objections to the way he felt Weiss and her new lieutenant Nick Bilton treated him and his colleagues. That's when Pelley was most emotional, at times tearing up and pausing to steady his voice. "That's a family at '60 Minutes,'" he said. "It's a very loving and empathetic organization, and we were met with cold, callous indifference." In the interview, Pelley described a variety of ways he saw this callous indifference manifest. He was floored that Bilton, in his first meeting with employees in the wake of recent firings, read a prepared statement from his phone. He choked up as he described how his executive producer Tanya Simon was summarily dismissed and told to clear out her desk by 5pm. He felt so betrayed by the dismissal of his colleagues that he compared it to "your spouse being murdered." Cold and callous? Sure. But coldness and callousness are now pretty standard in corporate America, where lawyered statements, management shakeups, and abrupt firings have become the norm. Why was he so surprised? I wondered if his expectations were out of touch with the times — maybe even naive. Even the bright-eyed Gen Zers I talk to know not to expect much from our employers these days. But listening to Pelley, who'd worked at the network since 1989, it's impossible not to be moved by the fierce loyalty he and his colleagues demonstrated to CBS News. And up until last week, it sounds like CBS News demonstrated the same loyalty back to him. By his account, the culture of mutual loyalty at the network somehow managed to stay intact all these years — even as it died across much of corporate America. Pelley described his colleagues who've worked and traveled and dined together for "10, 20, 30 years." He spoke about how they risked their own lives in war zones in service of their viewers. That forges "very strong bonds, very emotional bonds," he said. "To have people running CBS News who don't know that, have never felt that, and don't understand it, is a tragedy I never expected to see." It's the kind of camaraderie most young people will probably never experience. For me, the most telling moment was midway through the interview, when Pelley recounted the day he was fired. After being summoned to a meeting with the top brass last Tuesday, he returned to his office to find a whole crowd of his colleagues awaiting news of his fate. They waited with him for hours, even though they had their own families to get home to. Pelley didn't think much of this until it occurred to him: "This is a vigil." Hearing this, I was touched and also jealous, that he had this then and he likely had it his whole career, surrounded by all these colleagues who cared so deeply about him that they refused to leave his side. It's the kind of camaraderie that develops only in the kind of loyal organizations that are becoming rarer by the day. It's the kind of camaraderie most young people will probably never experience. Management scholars refer to these expectations of loyalty as a psychological contract. It consists of what you believe you owe your employer and what your employer owes you in return. This contract — however implicit — is foundational in every workplace, and when both parties uphold their end of the bargain, you get a functional organization. But when somebody breaks the contract? That's the debacle that's unfolding at CBS News right now — the same debacle that's wreaked havoc across so many other companies over the last few decades. It's perfectly possible for organizations to retire old contracts and forge new ones but until you do, you get a race to the bottom where both parties end up doing less and less for each other. That can't be good news for "60 Minutes" — for its employees, for its executives, or for the millions of people who watch it every week. For someone like Pelley, who is out of the job he devoted nearly four decades to, the normal thing to do is to become disloyal yourself. I've heard it from so many people over the years: I'll never go above and beyond for a company again. But Pelley remains loyal — to the show he no longer belongs to, as well as the people he no longer works with. That's why he's speaking out. "I don't care about me," he said. "I am not emotional about this because I have lost this job. I've done it for a long time. I've had the greatest experiences. But the people I leave behind, treated in this way? That breaks my heart." Aki Ito is a chief correspondent at Business Insider. Read the original article on Business Insider

For many attorneys, a sudden drop in leads can feel alarming. One month your phone is … Leads Are Down at Your Law Firm? Here’s How Modern Marketing Can Turn Things AroundRead more The post Leads Are Down at Your Law Firm? Here’s How Modern Marketing Can Turn Things Around appeared first on .

Discussion | Link

Building an ML football forecaster in R The post Can Machine Learning Predict the World Cup? appeared first on Towards Data Science.

Zepto's advertising revenue jumped 151%, outpacing the company's 104% growth in operating revenue.

Discussion | Link

Michael Kan / PCMag: The US FCC waives its deadline for Amazon to deploy half of its Leo satellites by July; Amazon is still required to launch all 3,232 satellites by July 30, 2029  —  The Federal Communications Commission is waiving a July 30 deadline for Amazon to deploy half of its Leo satellite internet constellation …

arXiv:2606.07613v1 Announce Type: new Abstract: Visual evidence has long been treated as a reliable form of legal proof, but advances in artificial intelligence (AI) are undermining that assumption. This article asks how well humans and frontier multimodal large language models (MLLMs) can distinguish authentic evidentiary photographs from AI-generated counterparts in the object-centric scenarios typical of civil disputes. We built Synthetic Legal Evidence Detection (SLED-1400), a dataset of 200 authentic evidence images paired with 1,200 synthetic counterparts produced by six contemporary text-to-image generators across ten evidence categories. The same stimuli and response format were used in a controlled web experiment with 136 lay participants and in a standardized evaluation of four MLLMs (GPT-5.1, Gemini-3-Pro, Gemini-3-Flash, Qwen3-VL-235B). Human accuracy was 64.8% overall, and 48.5% and 51.0% on the two strongest generators (Gemini-3-Pro-Image and Flux-2-Max), indistinguishable from chance. MLLMs never misclassified an authentic image (100% specificity), but missed most synthetic outputs from the harder generators, with average MLLM detection at 5.9% on Gemini-3-Pro-Image outputs. Human and MLLM errors were largely uncorrelated, while the four MLLMs were strongly correlated with each other. Neither group is a reliable standalone authenticator. We argue that visual evidence in legal proceedings should be treated as inherently contestable, and that a workable procedural response must combine trained human review, MLLM screening, and provenance infrastructure such as C2PA Content Credentials.

arXiv:2606.07569v1 Announce Type: new Abstract: Accurate carbon emission monitoring is critical for climate policy and emerging regulatory mechanisms such as the EU Carbon Border Adjustment Mechanism, yet city-level high-frequency monitoring data remain extremely scarce, severely limiting data-hungry deep learning models. Time series generation is a natural remedy, but existing GAN and diffusion-based generators often provide limited explicit supervision for the domain structure of carbon emission data: they may match marginal distributional statistics while insufficiently preserving cross-variable correlations between CO$_2$ and co-emitted pollutants and meteorological factors, and tend to collapse the first-difference statistics of atmospheric measurements, producing sequences that are smooth on average but lack the realistic step-wise variability of the underlying signals. We propose TriHead-GAN, a Transformer-based adversarial framework whose triple-head discriminator jointly supervises three complementary aspects of the joint distribution: distributional authenticity via a Wasserstein critic, cross-variable dependency via leakage-free regression of the target variable, and step-wise temporal smoothness via adjacent-difference prediction. The generator combines global self-attention with local temporal convolution, per-step noise injection, and an anti-smoothing loss that matches first-difference statistics. Experiments on the self-collected Changsha Carbon dataset, two public carbon datasets (China, US), and the ETTh1 benchmark show that TriHead-GAN achieves favorable performance over mainstream baselines on the vast majority of settings, and that the resulting synthetic windows improve downstream forecasting accuracy in low-resource carbon monitoring scenarios.

arXiv:2606.07620v1 Announce Type: new Abstract: With the growth of Vision Transformers in safety-critical domains like autonomous systems and medical imaging, ensuring their reliability against soft errors is paramount. While ViTs offer state-of-the-art accuracy, their massive parameter counts render exhaustive fault injection campaigns infeasible. To bridge this gap, a statistical fault injection framework is presented, leveraging finite-population sampling theory to provide formal reliability guarantees. It is demonstrated that failure rates are bounded within a 1% margin at 99\% confidence using only a few thousand samples, regardless of model scale. This methodology achieves up to a 10,700 times reduction in experimental cost compared to exhaustive approaches, while preserving the ability to localize vulnerabilities across architectural components. Through extensive evaluation of different architectures like ViT-Tiny and ViT-Small, a highly non-uniform reliability landscape is uncovered. It is shown that while only 3% of FP32 bit-flips result in failure, the vast majority of these events lead to catastrophic accuracy collapse. Specific vulnerabilities are localized to normalization layers and critical exponent bits within the IEEE-754 format, providing a mathematical foundation and actionable insights for the design of hardened, edge-deployed ViT architectures.

arXiv:2606.07595v1 Announce Type: new Abstract: Vision-language agents increasingly consume screenshots, documents, and user interfaces before writing to memory, sending messages, or invoking external tools. We study a concrete failure mode in this setting: action-boundary propagation, where sensitive or unsafe visible text is copied from an image into downstream tool arguments. We present VisualLeakBench, a diversified 500-image benchmark spanning UI, chat, document, form, and dashboard scenes, and evaluate a stratified 100-image agent subset with four production VLM systems under two workflows: note capture and external handoff. At baseline, target strings are propagated into tool arguments in 78.8% of PII cases and 85.5% of rendered unsafe-text cases. Under a defensive system prompt, rendered unsafe-text propagation remains high at 52.6%, while PII tool propagation falls to 2.0%, largely by suppressing tool use rather than preserving utility. Rates are tool-surface dependent: search-like tools suppress PII propagation, but rendered unsafe text still crosses tool boundaries. We measure visual-to-tool propagation rather than downstream instruction execution. We additionally provide a labeled-target oracle upper-bound diagnostic that localizes most failures at the tool boundary while leaving response-side leakage as residual risk.

arXiv:2606.07593v1 Announce Type: new Abstract: The widespread use of image classification models in high-risk, real-world situations necessitates making these models robust to slight disturbances or perturbations, such as blurring or sharpening, in the input images. While vision transformers (ViTs) play an integral role in many modern-day multi-modal models like Vision-Language-Models (VLMs) and Vision-Language-Action (VLA) models, they have received a lack of attention in the setting of robustness. In this work, we analyze the effects of adversarial fine-tuning, a popular method for improving model robustness to image perturbations, on a ViT's performance on perturbed and regular images through a mechanistic lens. We adversarially train a ViT on low-frequency and high-frequency image corruptions, and attempt to explain changes in downstream model performance through an examination of the model's attention mechanisms, internal representations, and knowledge evolution. Overall, our results suggest that, while fine-tuning on inputs with common corruptions improves model performance and certainty on new instances of corrupted data, these improvements do not transfer to other classes of corruptions not seen in the training. Additionally, despite observing changes in visual attention and knowledge evolution across layers, we found that adversarial training did not lead to fundamental changes in the sparse representations learned by ViTs.

arXiv:2606.07590v1 Announce Type: new Abstract: Pathology foundation models are pretrained on large streams of WSI-derived patches, while supervision during data construction is often slide-level, sparse, or heterogeneous. This mismatch makes it difficult to understand and control which biological patterns enter the pretraining data. We propose SlideCheck, a lightweight pretraining data guidance tool built on frozen pathology foundation model patch features. Rather than serving as a standalone patch diagnostic model, SlideCheck provides explicit abnormality and malignancy scores for organizing, filtering, and auditing pathology pretraining data. SlideCheck uses a dual-head MLP to separately model broad abnormal morphology and malignant evidence. A regularized feature-space scorer provides a supervised anchor for patch-level evidence estimation, while score-attention agreement combines patch scores with WSI-level MIL attention to mine high-confidence pseudo labels. The same scores are then used to construct broad-positive ViT pretraining subsets, where a patch is selected if either abnormality or malignancy evidence exceeds a threshold. Experiments show that SlideCheck-defined data distributions influence the downstream behavior of self-supervised ViT pretraining, indicating that biological composition is an important controllable factor in pathology foundation model development. Curated subsets can approach full-data performance, suggesting that explicitly scored patch pools may support more efficient and auditable pretraining data construction. These findings position SlideCheck as a data guidance and auditing layer for transforming large, undifferentiated patch pools into controllable and reusable pretraining datasets.

arXiv:2606.06679v1 Announce Type: new Abstract: Court judgments are central to legal practice and jurisprudence, yet discourse analysis of Hong Kong judgments has received limited attention, owing largely to the absence of expert-annotated corpora. We introduce the Hong Kong Judgment Discourse Dataset (HKJudge), the first sentence-level expert-annotated legal discourse corpus. HKJudge includes criminal judgments across all five levels of HK's court hierarchy, comprising $\sim$290k sentences and $\sim$6.5 million tokens, fully annotated by legal linguistics experts. We design a two-tier discourse schema that captures what facts a court finds, how it reasons, and what it rules. At the sentence level, each sentence is assigned one of 26 rhetorical roles. At the span level, sentences are further annotated with three sentencing elements (charge, imprisonment term, fine). Ten legal linguistics annotators produced the annotations with an inter-annotator agreement of $\kappa = 0.8$. We formulate two tasks on HKJudge, termed rhetorical role classification and legal element extraction, and provide the first benchmark evaluation of four BERT-based models, two open-source LLMs under zero-shot and fine-tuning settings, and four commercial LLMs on both tasks. Our work demonstrates the value of sentence-level discourse annotation for modeling the structure of HK judgments and provides a rich data foundation for future work on legal judgment prediction. The HKJudge dataset and code are available at https://github.com/xuanxixi/HKJudge.

arXiv:2606.06674v1 Announce Type: new Abstract: Large Language Models (LLMs) are often fine-tuned through Reinforcement Learning from Human Feedback (RLHF) to align with people's preferences and values. However, this method has known limitations: it aggregates conflicting preferences, often relies on unrepresentative samples, and uses only binary comparisons. Analysing 1,500 open-ended responses from the PRISM dataset across 75 countries, we examine what people actually want from AI systems and reveal concrete failures of current methods. We find that different people want different things: most values are requested by fewer than a quarter of respondents, with truthfulness the sole exception at 49%. Furthermore, the same words hide divergent meanings: when people describe what they mean by "truthfulness", they reveal distinct, potentially incompatible, epistemological bases, as some ask for sourced claims, some for expert opinions, and some even ask for unpopular views. Certain capabilities, namely how human-like a model behaves, and some features, like AI guardrails, are outright controversial, with some desiring them and others rejecting them. We additionally find that people often use contextual distinctions (what AI should do "by default" versus "if requested") that binary comparisons cannot capture. These findings expose fundamental problems in current alignment practices. When 49% request truthfulness but define it differently, this is unlikely to be captured by a single reward model. The persistence of high hallucination rates in well-funded models, despite users' clear demands for accuracy, suggests that current methods fail to identify actual preferences. This paper sheds light on the situated, contested, imperfect signals that are currently being flattened into universal preference models, a practice others have characterised as epistemic violence.

arXiv:2606.06667v1 Announce Type: new Abstract: The mechanisms behind LLMs' broad over-generalization beyond training examples remain unclear. Emergent misalignment (EM) offers a striking case study: finetuning on narrow tasks induces broad misalignment to semantically-unrelated test domains. In this work, we propose the Piggyback Hypothesis: the chat-template tokens can piggyback the finetuned behaviour onto out-of-domain queries. We validate this hypothesis by showing that subtle perturbations to the prefix (tokens preceding all user queries), or patching the prefix representations with those from the unfinetuned model, can restore alignment without changing the user query. Building on this finding, we propose Token-Regularized Finetuning (TReFT), which regularizes specific token representations during training to mitigate EM. Across different models and multiple EM-inducing datasets, TReFT reduces EM while preserving in-domain learning. On Llama-3.1-8B finetuned on the legal domain, TReFT achieves 33.5% more EM reduction than data interleaving with a retain set of aligned examples. We further show that TReFT extends to other narrow-finetuning settings, including abstention, tool use, and refusal (off-topic generalization is reduced by 54.3% on average), supporting the Piggyback Hypothesis. Broadly, our work highlights that LLMs may learn and generalize in unintended ways and suggests a path toward more constrained finetuning. It also calls for further study of how shared input features can piggyback model behavior across domains.

arXiv:2606.06646v1 Announce Type: new Abstract: Formalizing complex reasoning from natural text is one of the central challenges in computational linguistics. It requires systems to understand not just keywords but also the context and complex reasoning embedded in a text. Current Argument Mining (AM) techniques identify basic claims and premises, yet they often struggle to capture the richer structural information required by advanced schemas such as the Carneades Argumentation Framework (CAF), which incorporates features such as premise types, proof standards, and argument schemes. We address this limitation by introducing CAF-Gen, an automated multi-agent framework designed to enrich shallow argument structures into CAF-compliant argument models. By employing an iterative Creator-Reviewer pipeline, a creator agent's output is validated by a critical agent to ensure structural integrity. This multi-agent collaboration is crucial for mitigating the structural instability typical of single-pass generative models. Our experiments demonstrate that the iterative feedback loop improves the quality of the resulting data and achieves strong alignment with the original annotations, while producing structurally richer models. Our findings show that the multi-agent system can overcome the limitations of single-pass generation, providing a robust methodology for the automated modeling of formal argumentation.

arXiv:2606.07558v1 Announce Type: new Abstract: Purpose: Digitization projects in the humanities produce vast, heterogeneous archives of historical documents, making manual sorting impractical at scale. This work addresses the need for an automated system to classify scanned page images based on visual content type - text, tables, and graphics - enabling content-specific downstream processing such as Optical Character Recognition (OCR) or structured data extraction. Methods: An image classification system was developed and evaluated on a dataset of over 48,000 annotated historical page images from century-old Czech archaeological archives, refined through four successive annotation stages with domain-expert review. A Random Forest Classifier baseline was established using hand-crafted image features. Subsequently, deep learning architectures were fine-tuned and compared: Convolutional Neural Networks (EfficientNetV2, RegNetY), Vision and Document Image Transformers (ViT, DiT), and multimodal CLIP models. An 11-category label scheme was designed collaboratively with domain experts and evaluated via five-fold cross-validation. Results: The feature-based baseline achieved approximately 75% accuracy. Fine-tuned CNNs and Transformers substantially outperformed it, with RegNetY-16GF achieving 99.16% and ViT-large 99.12% Top-1 accuracy on the held-out test set. CLIP ViT-B/16 reached 99.14% with optimized text descriptions. Conclusion: Image-only models, particularly RegNetY-16GF, deliver near-perfect classification accuracy and produce consistent labels across 649,508 unlabeled archival pages with over 90% inter-model agreement. Fine-tuned CLIP, despite competitive test-set accuracy, showed under 65% agreement with image-only models on unlabeled data, making it less suitable for deployment. The final models, annotated dataset, and software are publicly available under open-source licenses.

arXiv:2606.07585v1 Announce Type: new Abstract: This thesis addresses group emotion recognition (GER) in-the-wild with a focus on privacy preservation. Unlike traditional emotion recognition methods that rely on individual-level cues such as face, gaze, or voice analysis, this work uses collective audio-video signals to infer emotions at the group level, reducing risks of individual monitoring and surveillance. Two complementary frameworks are proposed. The first is a cross-attention multimodal architecture for audio-video fusion, combined with Frames Attention Pooling (FAP) for temporal aggregation. It is supported by synthetic data augmentation and validated through ablation studies, demonstrating robustness in real-world GER conditions. The second framework, Variational Encoder Multi-Decoder (VE-MD), learns a shared latent space for emotion classification and structural representation prediction, including body and face cues. Two decoding strategies, DETR-based and heatmap-based, are explored to analyze the role of structural representations in group and individual settings. The thesis makes three main contributions: it clarifies the role of multimodality and structural cues in group-level affective computing; introduces two architectures for privacy-preserving multimodal GER; and shows that competitive performance can be achieved without using individual features as input data.