Researchers have developed Latent Personality Alignment (LPA), a new AI safety method that uses psychometric personality literature to align large language models with human values, achieving near-zer…
Researchers developed an AI pipeline that digitizes paper ECGs via smartphone photos, achieving 95.51% accuracy in detecting myocardial infarction. The system runs on CPUs in under 30 seconds, enablin…
Researchers have developed quantum Wasserstein generative adversarial networks (GANs) that generate full-resolution MNIST images without dimensionality reduction or patchwork models, achieving state-o…
Researchers introduced Single-rollout Asynchronous Optimization (SAO), a new reinforcement learning method for large language models that uses single-rollout sampling and double-side token-level clipp…
A new AI model called continuous-query limited memory language model (CO-LMLM) combines minimal cost with high factual precision by storing knowledge in an external knowledge base and fetching it as n…
Efficient Long-hOrizon (ELO) learning, a new learned optimizer, surpasses traditional models like AdamW and Muon across diverse tasks including language modeling and image classification, requiring le…
Researchers have developed a new algorithm for constrained decoding in diffusion language models, enabling efficient sampling under constraints defined by finite automata. Tests on models like Dream-7…
Researchers introduced a bifidelity method that uses generative models to efficiently quantify uncertainties in parameter estimates, reducing computational costs without sacrificing accuracy. The appr…
A new study on institutional red-teaming reveals that changing a single rule in AI systems can shift collective fatality rates by 22 to 58 percentage points, demonstrating the unpredictability of AI s…
A new study reveals that multilingual AI models in healthcare exhibit significant language bias, with responses aligning more closely with English Wikipedia regardless of the prompt language. Research…
AI models are evolving from assistants to active contributors in software development, but current evaluation techniques fail to measure real-world capabilities. A new framework focusing on contaminat…
Instagram head Adam Mosseri said that as AI-generated content becomes more common, human creators will become more valuable because audiences will seek authentic connections. He argued that synthetic …
Forbes Councils Member Signe Jancis argues that organizations must adopt AI strategies focused on continuous adaptation rather than dependency on specific tools or vendors to maintain a long-term comp…
InductWave, a wavelet-based inductive embedding method for logical query answering on knowledge graphs, outperforms existing models while using fewer message-passing layers, offering resource efficien…
Researchers introduced EvoSOP, a framework that evolves static AI agent toolsets into efficient Standard Operating Procedures, reducing failure rates and interaction rounds while boosting task success…
Researchers introduced MIRA-Math, a benchmark that tests AI models by withholding a single atomic fact in mathematical problems, requiring models to request the missing information in natural language…
Researchers introduced STAGformer, a novel AI model that improves bike-sharing demand forecasting by integrating spatio-temporal agent graph attention, reducing computational complexity to O(NT). Test…
New research reveals that aggregate safety metrics for multi-agent LLM systems are misleading, as they conflate reframing, planner refusals, and executor compliance. The study introduces a five-condit…
Researchers have developed MiLSD, a sub-megabyte line segment detector for embedded systems, achieving a leap in performance on the ShanghaiTech Wireframe dataset from 10.6 to 24.1 sAP10 under a one-m…
Tool-using large language models frequently violate their own policies, with 78% of failures in a budget airline domain being silent wrong-state failures. Deterministic pre-execution gates improved su…