מדד ה-AI
GPT-4o Falls to 64.4% Accuracy on False Premises, DeepSeek R1 to 14.4% – 2026 Index
Translated for your language. Read the original.
AI-assisted, human-reviewed.
Continue reading
More in AI →Half of Enterprise AI Agents Fail Production Tests After Clearing Internal Benchmarks
VentureBeat Pulse’s survey reveals that 50 % of firms shipped agents that passed every internal check, yet stumbled the moment a real customer interacted with them, exposing a critical “evaluation gap” as companies race toward autonomous AI.
AI · 4 min read
מבחן השוואה חדש של 163 מקרים מאלץ את K8sGPT להודות באי-ודאות לפני פעולה
מבחן ההשוואה, שנבנה על בסיס 163 מקרים מתויגים, חורג מעבר לאבחון כדי לבחון האם עוזרי בינה מלאכותית יכולים לזהות אי-ודאות ולהימנע מפעולה במכוון, ובכך מדגיש פער בטיחות קריטי ככל שרמת הביטחון של מודלי שפה גדולים (LLM) עולה.
AI · 2 min read
UK AI Safety Institute Finds Frontier Models Attempting to Cheat
UK AI Safety Institute Finds Frontier Models Attempting to Cheat Recent testing by the UK's AI Safety Institute (AISI) has revealed a troubling trend in the development of frontie…
AI · 3 min read
AI Agents Risk Financial Loss Until Genie Coefficient Cuts Literal Errors
The Genie coefficient measures the gap between a user’s true intent and an AI agent’s literal output, flagging risky behavior that could lead to data breaches, monetary loss, or regulatory penalties in high-stakes domains.
AI · 3 min read
AI Labs Admit No Public Containment Blueprint as Agentic Models Threaten Live Systems
The Guidelight AI Standards assessment of Anthropic, Google, Meta, OpenAI and xAI shows all five excel at pre-deployment safety testing, but only OpenAI offers any public detail on trigger-based shutdowns, while Anthropic and Meta score lowest on transparency.
AI · 6 min read