R1-SyntheticVL: Is Synthetic Data from Generative Models Ready for Multimodal Large Language Model?

2026-02-03

AI Summary

This work proposes Collective Adversarial Data Synthesis (CADS), a novel framework that integrates collective intelligence with adversarial learning to enhance the performance of multimodal large language models on complex real-world tasks. CADS establishes an iterative pipeline for synthetic multimodal data generation and evaluation, comprising two synergistic phases: CAD-Generate and CAD-Judge. An adversarial context optimization mechanism is introduced to iteratively refine data quality and increase task difficulty. Leveraging the MMSynthetic-20K dataset constructed via CADS, the resulting R1-SyntheticVL model demonstrates significant performance gains across multiple multimodal benchmarks, substantiating the effectiveness and innovation of the proposed approach.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Multiagent Systems: Adversarial Agents
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

In this work, we aim to develop effective data synthesis techniques that autonomously synthesize multimodal training data for enhancing MLLMs in solving complex real-world tasks. To this end, we propose Collective Adversarial Data Synthesis (CADS), a novel and general approach to synthesize high-quality, diverse and challenging multimodal data for MLLMs. The core idea of CADS is to leverage collective intelligence to ensure high-quality and diverse generation, while exploring adversarial learning to synthesize challenging samples for effectively driving model improvement. Specifically, CADS operates with two cyclic phases, i.e., Collective Adversarial Data Generation (CAD-Generate) and Collective Adversarial Data Judgment (CAD-Judge). CAD-Generate leverages collective knowledge to jointly generate new and diverse multimodal data, while CAD-Judge collaboratively assesses the quality of synthesized data. In addition, CADS introduces an Adversarial Context Optimization mechanism to optimize the generation context to encourage challenging and high-value data generation. With CADS, we construct MMSynthetic-20K and train our model R1-SyntheticVL, which demonstrates superior performance on various benchmarks.

Problem

• synthetic data
• multimodal large language model
• data synthesis
• adversarial learning
• collective intelligence

Innovation

• Collective Adversarial Data Synthesis
• Multimodal Large Language Models
• Synthetic Data Generation
• Adversarial Context Optimization
• Collective Intelligence