Understanding neural networks through sparse circuits
来自 OpenAI News
· 2025-11-13
OpenAI is exploring mechanistic interpretability to understand how neural networks reason. Our new sparse model approach could make AI systems more transparent and support safer, more reliable behavior.