JoshonSmith

https://artificialanalysis.ai/text-to-image/arena?tab=Leaderboard

考虑支持 LitePPO or ‘reinforce++ baseline’

1

**Describe the feature** LitePPO： 1、鲁棒的优势归一化 (Robust Advantage Normalization): 该技术结合了组级别和批次级别的统计量。具体来说，优势的均值在组（group）级别计算，而标准差在整个批次（batch）级别计算。 2、token级损失聚合 (Token-Level Loss Aggregation): 与DAPO中采用的技术类似，LitePPO主张在计算总损失时，对批次内所有token的损失进行求和，然后除以token总数，而不是在序列之间取平均。 **Paste any useful information** https://arxiv.org/abs/2508.08221

[Question] release the code of Generative RLHF-V

1

### Required prerequisites - [x] I have read the documentation . - [x] I have searched the [Issue Tracker](https://github.com/PKU-Alignment/align-anything/issues) and [Discussions](https://github.com/PKU-Alignment/align-anything/discussions) that this hasn't already been reported. (+1 or comment...