LLMs Trust Falsehoods Despite Warnings
New fine-tuning tests reveal LLMs confidently repeat false claims even after explicit warnings.
📂 Research
Latest articles in Research
LLMs Ignore Warnings, Hallucinate Truth
New fine-tuning tests reveal LLMs persist in stating falsehoods despite explicit warnings, highlighting critical alignme…
CVPR 2026: AI Medical Imaging Shifts to Workflow
Medical AI moves beyond accuracy metrics to adapt to complex clinical workflows and multi-modal data integration.
CVPR 2026: Medical AI Moves Beyond Image Recognition
CVPR 2026 highlights a shift in medical AI from simple image recognition to complex workflow integration and multi-modal…
NVIDIA Polar Boosts AI Coding Agents
NVIDIA introduces Polar, a token-faithful framework for GRPO training that enhances coding agents like Qwen and Claude w…
Peking U, Qiyuan, JD Launch RealAppliance for Robots
Peking University and partners release RealAppliance dataset to solve home robot planning issues via high-fidelity simul…
AI Scientist Robin: 2 Hours vs. 4 Months
Robin AI completes months of drug discovery work in under 2 hours, identifying a new treatment for blinding eye diseases…
Noroboto: Rust's Font Deception Risk
New research reveals 'lying fonts' in Rust AI tools, exposing critical visual deception risks for developers.
Noroboto: Rust's Lying Font Crisis
New 'Noroboto' vulnerability exposes font rendering risks in Rust apps. Developers must patch immediately.
AI Chatbot Uses 'Cognitive Inoculation' to Fight Health Misinfo
University of Oulu researchers developed an AI chatbot that uses cognitive inoculation to help users resist health misin…
AI Chatbot Fights Health Misinfo
University of Oulu researchers develop AI chatbot using 'cognitive inoculation' to combat health misinformation effectiv…
AI Accelerates Brain Drug Discovery
New AI models cut drug search time from decades to years, targeting MND treatments.