New method could increase LLM training efficiency

MIT News | Feb. 26, 2026, midnight | Original
By leveraging idle computing time, researchers can double the speed of model training while preserving accuracy.
Tackling industry’s burdensome bubble problem Designing a more resilient future for plants, from the cell up