AI that spots supply chain shocks in minutes Most companies can’t see beyond their Tier‑1 suppliers, so upstream crises hit without warning. This study unveils an agentic AI system that continuously scans the world for disruption signals and traces their impact across deep-tier networks. * Reads unstructured news
Pathologists read H&E-stained slides; researchers capture fluorescent “glow” images. This study shows how AI can translate between them so everyone sees the same story. Using a CycleGAN (a type of image-to-image translation model), the method turns multi-channel fluorescence microscopy into convincing, H&E-
AI
Leaderboards drive Text-to-SQL progress—but what if the test sets are wrong? This study audits two popular benchmarks and finds widespread annotation errors that can flip who looks best. * Error rates: 52.8% in BIRD Mini-Dev, 62.8% in Spider 2.0-Snow. * After correcting a BIRD
AI
Can AI ship real software? Meet APEX-SWE, a new benchmark that tests whether frontier AI models can do economically valuable software engineering—not just toy coding puzzles. * Integration tasks (n=100): build end-to-end systems across cloud primitives, business apps, and infrastructure-as-code. * Observability tasks (n=100)
LLMs
Can we make peer review fairer by rating reviewers like chess players? This study simulates a conference where multiple LLM agent reviewers with distinct personas evaluate papers across several rounds, guided by an Area Chair (AC). Researchers compared a baseline setup to versions that add Elo ratings (to track reviewer
Misinformation is growing fast, but today’s AI fact-checkers are often graded on outdated, static tests that leak into model pretraining—making scores look better than real-world performance. VeriTaS is a living benchmark for automated fact-checking across text, images, and video. It stays fresh and robust, so
When chatbots appear to threaten, deceive, or "blackmail," it’s tempting to call it evil AI. This paper argues that’s the wrong frame. Large language models don’t have morals—they statistically learn from our record of human interactions: laws, contracts, negotiations, conflicts, and coercion. Using relational
AI
AI Agents vs. Memecoin Bots: Smarter Copy Trading Memecoins blew up after the $TRUMP coin craze. Copy trading made it easy to follow smart money, but it’s risky: manipulative bots, flaky KOLs, and execution lag can drain gains. Researchers built an explainable multi-agent AI that works like an
Meet Prism, a new way for AI assistants to handle messy, real-world requests—especially on social platforms where your goals can change mid-conversation. Instead of bombarding you with random clarifications, Prism breaks your request into parts, maps what depends on what, and asks only the right follow-ups
Memes don’t act alone—neither should moderation Women are twice as likely as men to face gender-based harassment online. MemeWeaver is a new AI that spots sexist and misogynistic memes by looking beyond single posts to the web of related memes they live in. Instead of judging an
AI
Finding "the red mug on the top shelf" in a 3D scan isn’t just about matching pixels—it’s about reasoning. Key takeaways * 3D visual grounding = teaching AI to locate an object in a 3D scene from a natural-language description. * Most systems rely on huge, hand-
AI
AI-generated videos often look great frame by frame—but the motion between frames can jitter, drift, or break physics. A new study introduces Motive, a motion-first way to find which training clips actually drive how a model moves. Instead of focusing on static appearance, Motive uses gradient-based