Tag
13 articles
This article explains the AI trust crisis and why building reliable, honest AI systems is crucial for people to use AI effectively.
OpenAI has found that around 30% of tasks in the popular SWE-Bench Pro benchmark are broken, leading the company to withdraw its endorsement of the test.
OpenAI's analysis reveals significant methodological flaws in SWE-Bench Pro, a popular coding benchmark, raising concerns about the reliability of AI model evaluations.
A new Reuters Institute report shows that 10% of people worldwide now use AI chatbots for news weekly, but only 4% regularly verify sources, highlighting a trust gap in AI-generated content.
AI startup Probably raises $9M to build more reliable AI systems that prevent hallucinations and factual errors, aiming for accuracy comparable to deterministic systems.
Learn what AI hallucination means, how it happens, and why it matters for users and professionals. This beginner-friendly explainer covers the key concept behind recent AI reliability concerns.
Anthropic has released Claude Opus 4.8, an updated AI model focused on improved honesty and accuracy. The new version aims to reduce AI hallucinations by better acknowledging uncertainty and avoiding unsupported claims.
Anthropic's Claude Opus 4.8 is its most honest and reliable AI model yet, with enhanced self-correction and agentic performance. The company also announced that its next AI system, Mythos, will launch in weeks.
This article explains the technical mechanisms behind hallucinations in large language models, why they occur, and their implications for AI reliability and trustworthiness.
AI analytics agents are delivering wrong answers due to lack of governance, not because models are too small. Organizations must implement better oversight to ensure accuracy.
Researchers at Sapienza University of Rome have found that hallucinations in large language models leave measurable traces in their computations, offering a new method for detecting false outputs.
Researchers from PSU and Duke University develop a framework to automatically identify which agent in an LLM multi-agent system causes task failures and when the failure occurs.
