STORY · MODELLER_
GLM-5.2: Favors knowledge-based learning over reinforcement learning
Zhipu AI presented GLM-5.2 and argues that knowledge-based learning (KL) is more effective than reinforcement learning (RL) for model improvement. This contradicts DeepSeek's earlier approach, which prioritized RL-based optimization.
WHY IT MATTERS
The competition over the best method for improving large language models affects how future models are trained. If GLM-5.2 demonstrates superior results with a KL focus, it could influence the industry's investment and design choices.
SOURCES
MACHINE-GENERATED SUMMARY This summary is written by machine from the sources below. We sort and explain — but we are a way into the field, not the final word. Check the source when something matters to you.