Selected:

Researchers discover a shortcoming that makes LLMs less reliable
December 1, 2025
Large language models can learn to mistakenly link certain sentence patterns with specific topics — and may then repeat these patterns instead of reasoning.