A Carnegie Mellon University team, working with researchers from Johns Hopkins University and Aurora Innovation, proposes combining pretraining, fine-tuning, and reinforcement learning into a single algorithm. According to the study, the method improves the trade-off between accuracy and computing power, as well as model generalization and readiness for continual learning. The experiments involved models with approximately 124 million and 760 million parameters. The researchers say the approach simplifies model-development infrastructure, but no notable commercial applications or market partnerships had been announced as of October 30, 2026.
Source: Read the original article
Disclaimer: this content is for information purposes only and is not financial advice. Cryptocurrencies are highly volatile: you may lose all of your capital. Always do your own research. Legal notice
