Ruyi2.5实现多模态模型一次训练多端部署，兼顾性能与隐私。
Ruyi2.5 Technical Report
- 统一架构联合训练不同规模模型，保证多端语义一致。
- 在通用多模态评测中媲美Qwen3-VL，隐私监控任务表现更优。
- 提出二值响应优化算法，加速强化学习微调，提速2-3倍。
我们提出Ruyi2.5，一个基于AI Flow框架构建的多模态家庭模型。将Ruyi2的“训练一次，部署多次”范式拓展至多模态领域，Ruyi2.5采用共享主干结构，在单一统一流程中联合训练不同规模的模型，确保所有部署层级间的语义一致性。基于Ruyi2.5，我们开发了Ruyi2.5-Camera隐私保护摄像头服务系统，其采用两级识别架构：边缘端模型通过信息瓶颈引导的不可逆特征映射对原始帧去标识化，云端模型则执行深层行为推理。为加速强化学习微调，我们进一步提出二值前缀策略优化（BPPO），通过二值响应选择减少样本冗余，并将梯度更新聚焦于响应前缀，相比GRPO实现2至3倍训练加速。实验表明，Ruyi2.5在通用多模态基准上达到Qwen3-VL水平，而Ruyi2.5-Camera在隐私受限的监控任务中显著优于Qwen3-VL。
原文摘要 · Abstract (English)
We present Ruyi2.5, a multimodal familial model built on the AI Flow framework. Extending Ruyi2's "Train Once, Deploy Many" paradigm to the multimodal domain, Ruyi2.5 constructs a shared-backbone architecture that co-trains models of varying scales within a single unified pipeline, ensuring semantic consistency across all deployment tiers. Built upon Ruyi2.5, Ruyi2.5-Camera model is developed as a privacy-preserving camera service system, which instantiates Ruyi2.5-Camera into a two-stage recognition pipeline: an edge model applies information-bottleneck-guided irreversible feature mapping to de-identify raw frames at the source, while a cloud model performs deep behavior reasoning. To accelerate reinforcement learning fine-tuning, we further propose Binary Prefix Policy Optimization (BPPO), which reduces sample redundancy via binary response selection and focuses gradient updates on response prefixes, achieving a 2 to 3 times training speedup over GRPO. Experiments show Ruyi2.5 matches Qwen3-VL on the general multimodal benchmarks, while Ruyi2.5-Camera substantially outperforms Qwen3-VL on privacy-constrained surveillance tasks.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。