Ruyi2.5 Technical Report

2026-03-17

AI Summary

This work addresses the challenge of balancing efficiency and privacy in multimodal models under cross-deployment semantic inconsistencies and privacy-sensitive scenarios. Building upon the AI Flow framework, the authors propose a family of multimodal models with a shared backbone, enabling “train once, deploy everywhere.” To preserve privacy, they introduce an information bottleneck–guided irreversible feature desensitization mechanism within an edge–cloud two-stage inference architecture. Additionally, they design a Binary Prefix Policy Optimization (BPPO) algorithm for reinforcement learning–based fine-tuning. The approach matches the performance of Qwen3-VL on standard multimodal benchmarks while significantly outperforming it in privacy-constrained surveillance tasks, with BPPO yielding a 2–3× acceleration in training speed.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Computer Vision: Multi-modal Vision
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Security and Privacy: Security and privacy of machine learning and AI applications
• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

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.

Problem

• multimodal AI
• privacy-preserving
• edge-cloud collaboration
• semantic consistency
• surveillance tasks

Innovation

• multimodal familial model
• privacy-preserving camera system
• shared-backbone architecture
• Binary Prefix Policy Optimization
• information-bottleneck-guided feature mapping