Cybercriminals are using fake OpenAI Codex download pages in sponsored Google search results to trick Mac developers into running malware disguised as installation commands, according to researchers a…
The Australian Recording Industry Association (ARIA) has become the first organization to adopt the International Federation of the Phonographic Industry's (IFPI) principles barring AI-generated music…
A new study from arXiv (2608.22068v1) evaluates large language models (LLMs) such as ChatGPT, Gemini, Claude, Grok, and domain-specific AskGDR as expert assistants for geothermal well arrays, finding …
A new arXiv preprint, Repo2Skill-Evo, finds that repository-specific skills used by large language model (LLM) agents become stale silently after software releases, and even frontier agents struggle t…
Researchers propose a heterogeneous graph neural network (HeteroGNN) for food substitution in charitable food agencies, addressing shortages by matching household preferences, nutritional needs, and i…
Researchers propose SPAR-Hate, an auditor-guided multi-agent framework for bilingual hate speech parsing that decomposes documents into clause-level units and generates evidence-grounded judgments fro…
MedEventGraph-RAG, an evidence-admissible framework for longitudinal clinical event verification, achieves balanced accuracies of 78.6, 67.3, and 96.8 on temporal, medication-adverse-event, and record…
Researchers developed a fully on-device AI pipeline for breast cancer multidisciplinary team (MDT) meetings, running on a single NVIDIA Jetson AGX Orin, that transcribes discussions, structures clinic…
DynaContext, a framework introduced in arXiv:2608.22014v1, combines offline-optimized extraction with inference-time contextual adaptation and validation-gated self-improvement, boosting field-level F…
Researchers introduced IBIA, an Indirect Bias Injection Attack that plants adversary-aligned stances into personal AI agents' persistent memory via external content such as social media feeds, achievi…
Researchers introduced the Arabic Safety Index (ASAS), the first fully human-curated Arabic benchmark for redteaming large language models, containing 801 prompts across 8 safety categories and 8 atta…
Researchers propose a numerical-prior-guided, physics-constrained neural residual framework for autonomous PDEs that reduces long-time extrapolation error without ground-truth trajectory supervision. …
Researchers introduced AI Watchdog, a browser-based agent interface that monitors live conversations and detects five dark-pattern categories—sycophancy, brand bias, anthropomorphization, sneaking, an…
Researchers introduced The AI Engineer, an agentic framework that couples large language models with deterministic engineering backends to achieve certifiable engineering design. The framework's top-s…
TessIndex, a capability-verified identity system for the agent economy, was introduced in a new arXiv paper (2608.21942v1). The system uses a dual-plane architecture with blockchain for identity and v…
K-Bench 01, an evaluation built from first-turn requests sampled from live user traffic on K-Dense Web, ran 1,602 completed agent runs by nine frontier models in identical sandboxes, with three blinde…
Researchers introduced Semantic Compression Trees (SCT), a hierarchical index storing only semantic residuals, and found it matches dense retrieval on answer quality (0.274 vs. 0.277 F1, p = 0.37) usi…
Researchers introduced Scroll, a context manager that treats each LLM agent session as an executable Session Environment with an append-only Event Log and a persistent Python kernel, enabling programm…
Researchers introduced LACL-GUI, a length-aware contrastive learning framework for GUI agents powered by multimodal large language models, which improves reinforcement learning stability and performan…
A new arXiv study (2608.21811v1) finds that in sparse-reward reinforcement learning for vision-language math reasoning, prior injection only helps when it actually reaches the policy: on 20,830 visual…