single mention · optimization
Behavioral regularization
Behavioral regularization is a set of penalty mechanisms added to RL training to improve stability and encourage efficient, reliable behavior.1 It penalizes excessive generated-token growth relative to successful rollouts in a group, and applies segment-level penalties for format violations and tool call errors.2
How this definition was written
Summarised by glm-5.3-flash from 1 source; every sentence is checked against the quote it cites.
ToimproveRLtrainingstabilityandencourageefficient, reliablebehavior, weintroducetwocom- plementary mechanisms
— MiMo V2 6 technical reportA group-relative length penalty curbs excessive token growth, while segment-level behavioral penalties address format violations and tool call errors
— MiMo V2 6 technical report
Filed under optimization › training stability