Large-scale study of curiosity-driven learning

In three linesOpenAI releases large-scale study on curiosity-driven learning. Results show models trained with intrinsic curiosity mechanisms develop more robust and generalizable representations, improving downstream task performance without additional supervision.

Summary generated by Claude — human-verified