Researchers at Brown University tested whether modern AI language models can tell if events are common, unlikely, impossible or nonsensical. They used mechanistic interpretability and found internal patterns that match human judgments in several open-source models.
Photo by Zach M, Unsplash
A research team found that a programming style called "vibe coding" is producing insecure code with help from generative AI tools. A new radar scans public vulnerability data and flags cases that show AI signatures or risky patterns.
New research shows the human brain anticipates upcoming language by grouping words into grammatical phrases rather than predicting only the next single word. Scientists used brain recordings, behavioral tests and LLM measures across languages.
Researchers studied how large language models (LLMs) handle safety and tested training methods to reduce unsafe outputs while keeping performance. They identified key challenges and a technique that preserves safety during fine-tuning.
Researchers used neuroimaging and two adult datasets to test the Network Neuroscience Theory. They found that general intelligence reflects system-level organization and coordination across large-scale brain networks, with implications for development, injury and artificial systems.
Researchers tested Google’s NotebookLM, which turns research papers into podcast-style audio. The summaries were engaging and clearer for teaching, but every audio overview contained mistakes, so the authors advise reading the original papers to check claims.
