AI News Archive: April 30, 2026 — Part 14
Sourced from 500+ daily AI sources, scored by relevance.
- Understanding Adversarial Transferability in Vision-Language Models for Autonomous Driving: A Cross-Architecture Analysis
Vision-language models (VLMs) are increasingly used in autonomous driving because they combine visual perception with language-based reasoning, supporting more interpretable decision-making, yet their robustness to physical adversarial attacks, especially whether such attacks transfer across differe...
- Toward Autonomous SOC Operations: End-to-End LLM Framework for Threat Detection, Query Generation, and Resolution in Security Operations
Security Operations Centers (SOCs) face mounting operational challenges. These challenges come from increasing threat volumes, heterogeneous SIEM platforms, and time-consuming manual triage workflows. We present an end-to-end threat management framework that integrates ensemble-based detection, synt...
- REBENCH: A Procedural, Fair-by-Construction Benchmark for LLMs on Stripped-Binary Types and Names (Extended Version)
Large Language Models (LLMs) have achieved remarkable progress in recent years, driving their adoption across a wide range of domains, including computer security. In reverse engineering, LLMs are increasingly applied to critical tasks such as function and variable name recovery and type inference. ...
- Tracking Conversations: Measuring Content and Identity Exposure on AI Chatbots
AI chatbots are becoming a primary interface for seeking information. As their popularity grows, chatbot providers are starting to deploy advertising and analytics. Despite this, tracking on AI chatbots has not been systematically studied. We present a systematic measurement of web tracking on 20 po...
- Claw-Eval-Live: A Live Agent Benchmark for Evolving Real-World Workflows
LLM agents are expected to complete end-to-end units of work across software tools, business services, and local workspaces. Yet many agent benchmarks freeze a curated task set at release time and grade mainly the final response, making it difficult to evaluate agents against evolving workflow deman...
- DEFault++: Automated Fault Detection, Categorization, and Diagnosis for Transformer Architectures
Transformer models are widely deployed in critical AI applications, yet faults in their attention mechanisms, projections, and other internal components often degrade behavior silently without raising runtime errors. Existing fault diagnosis techniques often target generic deep neural networks and c...
- LLM-as-a-Judge for Human-AI Co-Creation: A Reliability-Aware Evaluation Framework for Coding
LLMs are increasingly employed both as judges for evaluating open-ended outputs and as co-creation partners in AI-assisted programming; yet rigorous evaluation in human-AI co-creation settings remains underdeveloped as judgments must be reliable, comparable across models, and interpretable over mult...
- Tail-aware N-version Machine Learning Models for Reliable API Recommendation
Machine learning (ML)-based API recommendation helps developers efficiently identify suitable APIs to complement the application code. However, code datasets used to train ML models often exhibit a long-tail distribution, leading to unreliable API recommendations, especially for infrequently used AP...
- Pragmos: A Process Agentic Modeling System
The advent of Large Language Models (LLMs) has significantly transformed tasks across Software Engineering. In the context of Business Process Management, LLMs are now being explored as tools to derive process models directly from textual descriptions. Existing approaches range from chatbot-driven s...
- To Diff or Not to Diff? Structure-Aware and Adaptive Output Formats for Efficient LLM-based Code Editing
Large Language Models (LLMs) are increasingly used for code editing, yet the prevalent full-code generation paradigm suffers from severe efficiency bottlenecks, posing challenges for interactive coding assistants that demand low latency and cost. Despite the predominant focus on scaling model capabi...
- Requirements Debt in AI-Enabled Perception Systems Development: An Industrial RE4AI Perspective
AI integration in automotive perception systems shifts requirements from static specifications to continuously evolving entities shaped by data, models, and operating contexts. When such changes are not consistently documented, validated, and traced, they accumulate as Requirements Debt (ReD), an un...
- The Grand Software Supply Chain of AI Systems
AI systems rest on software with low integrity mechanisms, leaving AI systems exposed across every stage from data acquisition to final inference. This paper makes the AI supply chain a first-class object of analysis, decomposing it across four architectural layers: data acquisition, model training,...
- PuzzleMark: Implicit Jigsaw Learning for Robust Code Dataset Watermarking in Neural Code Completion Models
Constructing and curating high-quality code datasets requires significant resources, making them valuable intellectual property. Unfortunately, these datasets currently face severe risks of unauthorized use. Although digital watermarking offers a post hoc mechanism for copyright authentication, exis...
- GenAI in Software Engineering: The Role of Technology Acceptance Models
Context: Many organizations are keen to incorporate generative~AI (GenAI) into their software development processes. Technology acceptance models, such as the Unified Theory of Acceptance and Use of Technology (UTAUT), are traditionally used to identify individual-level barriers to the acceptance of...
- LRS-VoxMM: A benchmark for in-the-wild audio-visual speech recognition
We introduce LRS-VoxMM, an in-the-wild benchmark for audio-visual speech recognition (AVSR). The benchmark is derived from VoxMM, a dataset of diverse real-world spoken conversations with human-annotated transcriptions. We select AVSR-suitable samples and preprocess them in an LRS-style format for d...
- BUT System Description for CHiME-9 MCoRec Challenge
Multi-talker automatic speech recognition (ASR) in conversational recordings remains an open problem, particularly in scenarios with large portion of overlapping speech where identifying and transcribing a target speaker is difficult from audio alone. Visual cues can help resolve speaker ambiguity, ...
- A Knowledge-Driven Approach to Target Speech Extraction in the Presence of Background Sound Effects for Cinematic Audio Source Separation (CASS)
We propose a knowledge-driven approach to speech target extraction in the presence of background sound effects already recorded in cinematic audio. The specific knowledge sources studied are manners of articulation that are detected in speech frames and adopted to form a knowledge vector as a part o...
- Predicting Upcoming Stuttering Events from Three-Second Audio: Stratified Evaluation Reveals Severity-Selective Precursors, and the Model Deploys Fully On-Device
Audio-based stuttering systems to date have been trained for detection -- what disfluency is present now -- leaving prediction, the capability needed for closed-loop intervention, unstudied at deployable scale. We train a 616K-parameter CNN on SEP-28k (Apple, 20,131 three-second clips) to predict wh...
- Position-Aware Drafting for Inference Acceleration in LLM-Based Generative List-Wise Recommendation
Large language model (LLM)-based generative list-wise recommendation has advanced rapidly, but decoding remains sequential and thus latency-prone. To accelerate inference without changing the target distribution, speculative decoding (SD) uses a small draft model to propose several next tokens at on...
- Purifying Multimodal Retrieval: Fragment-Level Evidence Selection for RAG
Multimodal Retrieval-Augmented Generation (MRAG) is widely adopted for Multimodal Large Language Models (MLLMs) with external evidence to reduce hallucinations. Despite its success, most existing MRAG frameworks treat retrieved evidence as indivisible documents, implicitly assuming that all content ...
- NuggetIndex: Governed Atomic Retrieval for Maintainable RAG
Retrieval-augmented generation (RAG) systems are frequently evaluated via fact-based metrics, yet standard implementations retrieve passages or static propositions. This unit mismatch between evaluation and retrieval objects hinders maintenance when corpora evolve and fails to capture superseded fac...
- SimEval-IR: A Unified Toolkit and Benchmark Suite for Evaluating User Simulators and Search Sessions
User simulators are increasingly central to interactive information retrieval, yet the community lacks standardized evaluation tools. Simulators serve two objectives, behavioral realism (matching real user behavior) and tester reliability (producing valid system rankings), and these are often confla...
- Reproducing Adaptive Reranking for Reasoning-Intensive IR
The classical cascading pipeline of retrieve--rerank suffers from a bounded recall problem, stemming from limitations of the first-stage retriever. Most current approaches address the bounded recall problem by improving the first-stage retriever, but this incurs substantial training and inference co...
- MethylBench: A comprehensive benchmark of DNA methylation profiling methods across diverse sequencing platforms
Background: DNA methylation can be profiled using multiple technologies that vary in resolution, coverage and cost. Yet systematic benchmarks across these methods remain scarce. Methods: We compared six widely used technologies - Illumina EPIC array, TWIST, Whole-Genome Enzymatic Conversion, Reduced Representation Bisulfite Sequencing, long-read genome sequencing (LR-GS) with Pacific Biosciences (PacBio) and Oxford Nanopore Technologies (ONT) - using Genome in a Bottle (GIAB) reference samples and ten samples derived of blood and fibroblast cultures of 5 individuals. We assessed CpG coverage, consistency of differentially methylated cytosine (DMC) detection and genomic annotation, with particular attention to overlapping signals across assays. Results: Despite major differences in assay design, all technologies consistently identified DMCs enriched in promoter and intronic regions, highlighting these loci as robust hotspots of epigenetic variability. Annotation redundancy strongly infl
- Automatic deep learning-based segmentation and quantification of stented arterial cross-sections for morphometric analysis
Arterial vascular diseases, such as atherosclerosis, are among the most serious global health threats. In preclinical studies, morphometric analysis of histological arterial cross-sections is considered the gold standard for assessing vascular remodeling and the effectiveness of therapeutic interventions. However, morphometric analysis is usually performed manually, which is time-consuming, subjective, and requires significant user interaction. This paper presents a fully automated, operator-independent framework for the precise morphometric analysis of stented arterial cross-sections, extending the previously developed qHisto (quantitative histology) framework for the quantification of various histological components. A neural network for the segmentation of arterial structures was trained and evaluated using 819 cross-sections. In addition, a quantitative analysis of vascular morphology, fibrin area, and lumen asymmetry was performed using 72 cross-sections from coated and uncoated b
- Systems Pharmacology Reveals Type I Interferon and Myeloid-Like B Cell Reprogramming as Druggable Axes in Antiphospholipid Syndrome
Antiphospholipid syndrome (APS) lacks targeted therapies beyond anticoagulation, and its molecular heterogeneity remains poorly characterized. We employed an integrative systems pharmacology approach combining weighted gene co-expression network analysis (WGCNA), single-cell RNA sequencing, Connectivity Map (CMap) screening, and molecular docking to identify druggable targets in APS. WGCNA of bulk RNA-seq data from neutrophils (n = 18) and whole blood (n = 88) identified two disease-associated modules: ME10 (176 genes, r = 0.77, interferon-I signaling) and ME2 (3409 genes, r = 0.79, degranulation/innate activation). Single-cell analysis of 26,936 B cells revealed transitional B cells with elevated ME2 scores and aberrant SPI1 expression, suggesting myeloid-like transcriptional reprogramming. CMap analysis ranked chloroquine, a first-line APS therapy, among top ME2 candidates (NCS = -2.07), validating the computational approach. DrugBank mapping identified 14 FDA-approved drugs targetin
- A Conditional Variational Autoencoder with QSAR-Guided Surrogate-Weighted Fine-Tuning and Cross-Entropy Optimization for Targeted Antimicrobial Peptide Generation
Machine Learning frameworks have emerged as a promising tool for antimicrobial peptide design; however, generative models remain limited by two persistent problems: the limited availability of experimentally validated peptides and the circular dependency of the models. In this work we present a conditional variational autoencoder pipeline that addresses both limitations through a modular architecture that combines both binary and quantitative experimental data and implements a multimodal approach to externally guide the generation. A transformer-based encoder successfully generated a discriminative 64-dimensional latent space (test AUROC 0.968, F1 0.919) separating antimicrobial from non-antimicrobial sequences. This latent representation conditions a species-specific LoRA fine-tuned ProtGPT2 decoder through a scalar gating function, which generates balanced antimicrobial peptides through two different modes; prior and perturb, depending on their generation starting points. We introduc
- Physics-informed self-supervised learning enables spectra-free multiplexed imaging on standard fluorescence microscopes
Multiplexed fluorescence imaging is limited by spectral overlap and the small number of excitation or emission channels available on standard microscopes, restricting most laboratories to low-plex imaging. Here we introduce physics-informed spectra-free multiplexed imaging (PhySMI), a self-supervised framework for underdetermined spectral unmixing that enables highly multiplexed imaging without dense spectral measurements after training. By embedding the spectral forward-mixing process into a self-consistent architecture, PhySMI recovers physically plausible source decompositions from unlabeled data without paired ground-truth labels while suppressing stochastic acquisition noise. PhySMI resolves five subcellular structures from only three excitation channels, overcoming the conventional channel-number limit while preserving spectral fidelity and minimizing crosstalk (<0.5%). The framework also generalizes across imaging systems, enabling zero-shot deployment on standard fluorescence m
- Species-specific transformer models of bacterial gene order and content for genomic surveillance tasks
Transformer models enable functionally meaningful representation of complex biological data, such as nucleotide or protein sequences. Existing foundation transformer models are trained on large multi-domain corpuses of unlabelled DNA or protein data, showing unmatched task generalisation. However, these foundation models are often outperformed on domain-specific tasks by models trained on taxonomically-constrained data, such as gene classification in prokaryotes. By extension, species-specific transformer models hold promise for targeted analyses, given sufficient training data are available. Epidemiological analysis of bacterial pathogens exemplifies the use-case of species-specific transformers, due to the wealth of genome data available, coupled with pathogen-specific analyses carried out during routine and outbreak surveillance. Here, we trained a transformer model, PanBART, on the gene content and gene order of two important and biologically distinct bacterial pathogens, Escherich
- Development and Validation of a Two-Stage NLP-LLM System for Automated Extraction of Deprescribing Recommendations from Discharge Summaries
Introduction: Polypharmacy in older adults is associated with increased risks of adverse drug events and functional decline. Discharge summaries often contain deprescribing recommendations, but these are frequently overlooked due to documentation complexity. Objective: To develop and validate a two-stage hybrid system combining rule-based natural language processing (NLP) and large language model (LLM) for automated extraction of deprescribing recommendations from discharge summaries. Methods: This retrospective cohort study included 850 discharge summaries from patients aged [≥]65 years with hospitalisation [≥]48 hours across six public hospitals in New South Wales, Australia. Model 1 (rule-based NLP) extracted discharge medications and candidate sentences containing pre-defined deprescribing keywords. Model 2 (open-source LLM) classified candidate sentences into five categories. Data were split into training (80%) and test (20%) sets. Gold standard classifications were established by
- Auditing What Was Said: The Epistemic Promise and Limits of Ambient AI in Clinical Practice
Traditional audit methods that rely on written records often miss the nuances of clinical reasoning that influence patient care. Ambient artificial intelligence captures spoken clinical encounters, allowing the analysis of real clinicianpatient dialogue at scale. In a study of 124 urology consultations, a transcript-centered audit identified inter physician variation and expert disagreement that conventional review missed. We explore the epistemic gains of this approach, its nonverbal blind spots, behavioral effects, technical vulnerabilities, and the EU AI Act's regulatory landscape.
- Comparing Physicians' Assessments of Context-specific AI-powered clinical reasoning assistant with General-Purpose AI agent: A Prospective Multi-Site Physician Evaluation of VITA versus ChatGPT in India and Bangladesh
Abstract Background: Healthcare providers in low- and middle- income countries (LMICs) are increasingly relying on Artificial Intelligence (AI) tools, yet most available AI assistants are general-purpose systems not designed for the specific clinical, epidemiological, and resource contexts of these settings. There is no evidence, from physicians' assessments, on whether clinical reasoning support from purpose-built, context-specific and retrieval-augmented AI tools can outperform general-purpose AI agents. Methods: We conducted a prospective multi-site validation study enrolling 37 physicians across India and Bangladesh. Each physician evaluated two AI tools (a) VITA (Validated Intelligence for Treatment and Assessment), a purpose-built (context-specific and retrieval-augmented) clinical reasoning AI assistant trained on India-specific guidelines, antimicrobial resistance patterns, and formulary constraints, and (b) ChatGPT Plus (version 5.2), a leading general-purpose AI assistant on
- Forecasting Minute-by-Minute Stress, Anxiety, and Affective States Using Time-Series Analysis of Wearable Sensor Data
This paper focuses on forecasting minute-by-minute stress, anxiety, and affective states using wearable sensor data. It addresses mental health as a growing concern and the limitations of traditional assessment methods. A time-series machine learning framework was developed using electrodermal activity (EDA) and heart rate variability (HRV) features from the WESAD dataset. Models were trained and evaluated for minute-by-minute prediction of self-reported psychological states. Both classification (stress, anxiety) and regression models (affect) were explored comparing time-series and static approaches. Findings support the feasibility of real-time, personalized mental health monitoring using wearable devices and their potential for timely interventions in clinical or remote settings. The paper demonstrates how temporal modeling can enhance emotional state prediction and inform future research and system development.
- Explainable AI Predicts Hematoxicity from Cancer Treatment Using Multimodal Real-World Data
Adverse drug effects remain a major barrier to safe and effective cancer therapy, underscoring the need for tools that predict treatment-related toxicities. We analyzed multimodal real-world data from 14,596 cancer patients across 38 cancer entities, encompassing 330 clinical, tumor, and imaging characteristics, along with 89 anticancer agents. Hematological adverse events (HAE), defined by nadirs of hemoglobin, leukocyte, neutrophil, and platelet values within two months of treatment initiation, were highly prevalent (87.7%; 33.1% severe). We developed Toxix, an explainable artificial intelligence (xAI) framework modeling interactions between patient characteristics and drug combinations. Toxix achieved strong predictive performance for severe toxicities (median AUROC 0.85 for anemia; >0.76 for leukopenia, neutropenia, and thrombocytopenia) and was validated in an external cohort of 2,768 patients with non-small cell lung cancer. Model explainability enabled systematic characterizatio
- A biologically annotated neural network for proteomic discovery in Parkinsons disease
Machine learning models that can utilize high-dimensional data to make predictions and derive biological insights can improve understanding of diseases. Here, we develop a biologically annotated neural network model for proteomics data (P-BANN) which has several practical advantages: (1) it incorporates known relationships between proteins and signaling pathways into its architecture design; (2) it uses Bayesian principles to enable variable selection on the most important proteins for a disease of interests; and (3) it combines structured and black-box variational inference to analyze different classes of phenotypes at scale. To demonstrate the value of the approach, we apply P-BANN to one of the most common neurodegenerative disorders: Parkinsons disease (PD). We consider two biomarker-defined phenotypes within the PD population: presence of neuronal-predominate aggregated alpha-synuclein in cerebrospinal fluid, and changes in dopamine transporter binding in the striatum on imaging.
- The New AI Leader? Anthropic’s $1 trillion signal stuns Silicon Valley
The New AI Leader? Anthropic’s $1 trillion signal stuns Silicon Valley YourStory.com
- Clinician Discourse on Ambient AI Scribes: A Reddit-based Topic Modelling and Sentiment Analysis
Background. Ambient AI scribes are rapidly entering clinical workflows, yet end-user perspectives remain underrepresented in the peer-reviewed literature. Online clinician communities offer an unfiltered window into adoption barriers, perceived benefits, and product-level concerns. Objective. To characterise themes and sentiment in clinician discourse on ambient AI scribes across professional Reddit communities. Methods. We scraped posts from ten clinically oriented subreddits using twelve AI scribe related queries via the public Reddit JSON API. A two-tier keyword filter retained posts mentioning at least one AI scribe term and one clinical or workflow term. Texts were embedded with all-MiniLM-L6-v2, reduced via UMAP, clustered with HDBSCAN, and labelled using BERTopic with c-TF-IDF keyword extraction. Noise topics matching predefined off-topic patterns (for example, residency match, finance) were removed. Themes were assigned concise labels via Claude Sonnet 4. Sentiment was classifi
- Sources: Anthropic potential $900B+ valuation round could happen within 2 weeks
Anthropic is asking investors to submit allocations for the AI company’s latest fundraise within the next 48 hours, according to sources familiar with the matter.
- Anthropic weighs new funding round at valuation exceeding $900 billion, Bloomberg News reports
Anthropic weighs new funding round at valuation exceeding $900 billion, Bloomberg News reports [Ads] Plant Trees with CharityTree Collect trees by shopping or using online services, and we’ll plant them with our partners for you! [ Install to Chrome ]
- Anthropic is eyeing a funding round that could make it bigger than OpenAI
The Claude maker's could emerge with a $900 billion valuation, making it bigger than ChatGPT maker OpenAI
- Anthropic weighs new funding round at valuation exceeding $900 billion: Report
Artificial intelligence firm Anthropic is exploring a new funding round. This round could value the company at over $900 billion. This valuation would make Anthropic the world's most valuable AI startup. The company is considering offers at more than double its current valuation. A decision on the round and its valuation is expected in May.
- Anthropic weighs funding offers at over $900 billion valuation, tops OpenAI
Anthropic had previously resisted several inbound proposals from investors for a new round at a valuation of $800 billion or more
- Anthropic reviewing investor offers that would value the company at over $900 billion
Anthropic is reviewing investor offers for a new funding round that would value the AI company at over $900 billion, Bloomberg reports. The article Anthropic reviewing investor offers that would value the company at over $900 billion appeared first on The Decoder .
- Claude maker Anthropic gets funding offers that would value it at $1.15 trillion, higher than OpenAI
Claude maker Anthropic gets funding offers that would value it at $1.15 trillion, higher than OpenAI The Straits Times
- Swedish legal AI startup Legora raises €42 million extension to Series D – bringing total to over €500 million
Stockholm’s Legora today announced a €42 million ($50 million) extension of its previously announced Series D financing, bringing the total round to €513 million ($600 million) in equity and valuing the company at €4.7 billion ($5.6 billion) post-money. The extension adds Atlassian and NVentures (NVIDIA’s venture capital arm) as corporate investors, alongside new financial investors, […] The post Swedish legal AI startup Legora raises €42 million extension to Series D – bringing total to over €500 million appeared first on EU-Startups .
- Swedish Legal Tech Startup Legora Lands Another $50M In Nvidia-Led Series D Extension
Swedish Legal Tech Startup Legora Lands Another $50M In Nvidia-Led Series D Extension Crunchbase News
- OpenAI announces new advanced security for ChatGPT accounts, including a partnership with Yubico
OpenAI is launching additional opt-in protections for ChatGPT accounts. The new security initiative includes a new partnership with security key provider Yubico.
- OpenAI now lets you lock your ChatGPT account with a hardware key. Here is why it thinks you should.
OpenAI has released a security feature for ChatGPT accounts that treats them the way banks treat online banking: hardware keys, no passwords, no email recovery, and no help from customer support if you lose access. The feature, called Advanced Account Security, is an opt-in setting that requires users to authenticate with two passkeys, two hardware […] This story continues at The Next Web
- Google’s Gemini AI assistant is hitting the road in millions of vehicles
The move signals Google’s push to bring more advanced, conversational AI into the driving experience.
- GM Adds Google Gemini for Drivers to Rev Up With AI Assistant
You must have a GM car from 2022 or newer, and already have the Google built-in operating system -- it can't be retroactively installed.