Microsoft warns that AI-driven vulnerability detection will lead to more frequent security patches for Windows. The company's new multi-model agentic scanning harness (MDASH) uses AI to scan code at s…
A Chinese technology company is set to become the largest shareholder in an AI agent start-up after Beijing blocked a U.S. acquisition, signaling China's strategic push to dominate the AI sector. The …
A new AI method using prefix control tokens with BART fine-tuning boosts ad headline click-through rates by 5.82% and Rouge-L scores by 25.82%, giving advertisers customizable control over headline ge…
Researchers have developed a new structured pruning method for large language models that borrows from unstructured Adaptive Feature Retention, addressing distribution mismatches, sign information los…
Anthropic has discovered a hidden facet within large language models that suggests greater complexity than simple algorithmic responses, reigniting debates about AI consciousness. The finding raises e…
Researchers developed a new method for out-of-scope intent detection using MiniLM embeddings, treating it as a one-class classification task with multi-cluster boundary learning. The approach outperfo…
DeepSearch-Evolve introduces a self-distillation framework for training web agents in a controlled environment, achieving state-of-the-art results on benchmarks like BrowseComp, GAIA, and HotpotQA wit…
Researchers have introduced a new causal inference framework called triangulation that combines estimates from multiple models with different assumptions to infer true causal effects, avoiding the nee…
PhasorFlow, an open-source Python library, introduces a lightweight computing paradigm based on the S^1 unit circle for AI tasks, offering efficient alternatives to traditional algorithms. Its innovat…
Researchers introduced DeepAbduction, a new dataset, and the Inverse-Forward Abductive Reasoning (IFAR) framework to improve abductive reasoning in large language models. The IFAR framework achieved a…
Researchers Sharp et al. introduced the concept of 'agentic inequality' in 2025, highlighting disparities in AI access. A new framework, the Context Access Divide (CAD), reveals how AI's ability to au…
VEGAS (Video caption Evaluation via GAze Score) introduces a training-free method to align video captions with viewer gaze data, improving caption-to-video retrieval and personalization. The approach …
Researchers applied a Spectral Temporal Graph Neural Network (StemGNN) to predict user equipment scheduling states in 5G networks, recovering 57-73% of sum rate loss caused by backhaul latency. The mo…
Researchers developed a Q-learning-based adaptive retraining method to address traffic drift in Open Radio Access Networks (O-RAN), reducing retraining overhead while maintaining forecasting accuracy.…
Researchers introduced WCog-VLA, a dual-level World-Cognitive Vision-Language-Action model that enables autonomous vehicles to transition from reactive to proactive navigation by integrating world cog…
A new multi-agent pipeline using large language models and knowledge graphs is transforming risk management in critical infrastructure by converting natural-language system descriptions into audit-rea…
An open-source firewall designed to secure Large Language Model interactions has been released, achieving F1 scores up to 94.93% in detecting data leaks and privacy threats. The framework uses a multi…
Researchers have developed VRAIN, a hybrid AI system that combines text-to-image generation with context-aware editing to improve rare-class instance segmentation. The approach achieves up to 4.0 poin…
Prismata introduces a new security standard for autonomous web agents by enforcing contextual least privilege, protecting them from prompt injection attacks without requiring developer intervention. T…
Researchers have developed NamedCurves+, an image editing tool that combines color naming and deep learning to enable intuitive, customizable enhancements. The system uses tone curves and transformer-…