Researchers found that eight AI models struggle with cultural context, failing to differentiate between cultural groups despite possessing latent knowledge. Disrupting key connections reduced cultural…
Meta launched an AI tool that lets users alter Instagram content, sparking backlash over concerns it blurs the line between genuine and fabricated posts, threatening user trust and enabling potential …
Researchers have developed HVR-Met, an AI system that improves extreme weather diagnostics by integrating expert knowledge and a novel feedback loop. The system sets a new benchmark for meteorological…
Researchers introduced ARLArena, a framework for stabilizing agentic reinforcement learning (ARL), and SAMPO, an optimization method that addresses ARL's instability. The tools aim to make ARL more re…
A new Transformer-based model for sign language recognition achieves 88.4% accuracy on unseen classes from the LSA64 dataset using only eight reference examples per class, demonstrating strong one-sho…
A new study demonstrates that discrete diffusion models, when combined with online fine-tuning, replay mechanisms, and validity penalties, outperform traditional methods in molecular optimization acro…
Researchers introduced ANTAP (Automatic Non-Textual Agent Picker), a new defense mechanism for multi-agent systems that replaces unreliable textual self-descriptions with empirical capability testing.…
Researchers introduced the Native Differentiable Virtual Machine (NDVM), a runtime structure that blends symbolic structure with numerical computation to accelerate AI model calibration. NDVM achieves…
A new benchmarking framework for mutual information estimation reveals that no single estimator works best across all scenarios, challenging the conventional wisdom of one-size-fits-all solutions. The…
Researchers at the SMM4H-HeaRD 2026 conference used deep learning models to predict cancer stages from pathology reports, with LightGBM and TF-IDF achieving high AUROC scores. However, performance dro…
Order-equivariant neural networks (OENN) are advancing AI by leveraging symmetry to improve deep learning efficiency and performance. Researchers have characterized linear order-equivariant maps, cons…
Researchers introduced SFT+RL, a two-stage framework combining supervised fine-tuning and reinforcement learning on CLIP's encoder, achieving a 10.2% improvement in clean accuracy and 15.8% in adversa…
Researchers introduced the Dynamic Security Control Compositor (DSCC), a two-phase system that composes tool security policies and tracks data sensitivity to secure multi-tool AI agents. The DSCC bloc…
Researchers developed an AI framework using semantic segmentation on H&E-stained images to diagnose liver cancer types, achieving high accuracy rates of up to 1.000 for colorectal metastatic adenocarc…
A review of 74 studies on AI in open-source intelligence (OSINT) finds that while large language models and agentic AI show promise for data collection and analysis, evaluation frameworks lag behind, …
Researchers introduced CONTRA, a new algorithm that uses large language models to identify configurations in autonomous agents that could lead to malicious actions. Testing on 473 popular skills found…
AI integration in security systems requires human oversight to ensure accountability and adaptability, as shared architectures create blind spots and systemic errors. A balanced division of labor betw…
Vision-language models exhibit a source-credibility bias, prioritizing logos and names over content, according to a study by the CueTrust initiative. The bias, measured via a Source-Override Index, wa…
AnchorVLA, a new planning framework for autonomous driving, uses trajectory-pattern anchors to bridge high-level reasoning and execution, achieving state-of-the-art results on the Bench2Drive benchmar…
New methods in AI aim to 'unlearn' specific data, enhancing privacy without sacrificing utility. The Marginal Unlearning Principle provides provable guarantees for data-point removal, while feature un…