Reinforcement Learning Finally Works in Latent Reasoning Space — But Only After Fixing Three Ways It Breaks

Reinforcement Learning Finally Works in Latent Reasoning Space — But Only After Fixing Three Ways It Breaks | BedrockNews