Ruyi2.5 Technical Report
Fuente:
arXiv
Salvato in:
| Autori principali: | , , , , , , , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2026
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866918395280097280 |
|---|---|
| author | Song, Huan Tian, Shuyu Zhao, Qingfei Hong, Wenhao Liu, Jiang Long, Ting Shao, Jiawei Li, Xuelong |
| author_facet | Song, Huan Tian, Shuyu Zhao, Qingfei Hong, Wenhao Liu, Jiang Long, Ting Shao, Jiawei Li, Xuelong |
| contents | 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. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2603_17311 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Ruyi2.5 Technical Report Song, Huan Tian, Shuyu Zhao, Qingfei Hong, Wenhao Liu, Jiang Long, Ting Shao, Jiawei Li, Xuelong Computation and Language 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. |
| title | Ruyi2.5 Technical Report |
| topic | Computation and Language |
| url | https://arxiv.org/abs/2603.17311 |