Training

Introduction Over the past month I’ve written two articles on decision models: Jev vs Laya compared the closed-source Jev with the open-source Laya, and Microsoft-Decision-1 Explained added Microsoft to the picture. The question I’ve been asked most since then is: can I train one myself? Yes, you can, and the bar is lower than you might think. Just look at the third-party leaderboard JevBench: as of mid-October, its open-weights board ranks more than 160 systems, the vast majority built by the community and small teams. The ones at the top aren’t mysterious either: H2O-Lightning-4B, first on the composite score, is a fine-tune of Alibaba’s open-source Qwen3.5-4B; Quyet-1.0-Large, first on the capability score, is Gemma-4-31B plus an adapter; and Microsoft’s own Microsoft-Decision-1 is also built on the open-source Qwen3.5-9B. ...

Introduction In the post-training stage of large language models, reinforcement learning (RLHF / RLAIF) has become one of the key factors that determines the upper bound of model capability. Recently, GLM-5.2 switched its training algorithm from the GRPO (Generalized Reward Policy Optimization) used in GLM-5.1 to the more classical PPO (Proximal Policy Optimization), bringing a clear improvement in results. This change is not a simple “algorithm replacement”. It is a systematic upgrade in stability, generalization, and training controllability. ...