BedrockNews

BedrockNews

Ctrl + K
Log In
Community
A Difficulty Thermostat for RL Training: AdaPrefix-GRPO Doubles Small-Model Math Accuracy by Feeding Just Enough Hints | BedrockNews