SynthRL: Scaling Visual Reasoning with Verifiable Data Synthesis
This work addresses the challenge of data scarcity for training vision-language models in visual reasoning, offering a scalable solution that enhances performance on complex tasks, though it is incremental in building upon existing RLVR methods.
The paper tackled the problem of scaling visual reasoning by proposing SynthRL, a pipeline for synthesizing verifiable and challenging data to improve reinforcement learning with verifiable reward (RLVR) training, resulting in models that achieved consistent gains across five out-of-domain benchmarks, with significant improvements over baselines, particularly on the most challenging samples.
Vision-language models (VLMs) trained via reinforcement learning with verifiable reward (RLVR) have shown notable progress in scaling test-time compute effectively. In this work, we investigate how synthesized RL data can further improve RLVR. To this end, we propose \textbf{SynthRL}-a scalable and guaranteed pipeline for automatic data scaling in reasoning-oriented RL training. SynthRL comprises three key stages: (1) selecting seed questions with appropriate distribution, (2) augmenting them into more challenging variants while preserving the original answers, and (3) a guaranteed verification stage that ensures near-perfect correctness and difficulty enhancement. Our empirical experiments demonstrate SynthRL's scalability and effectiveness. When applied to the MMK12 dataset, SynthRL synthesizes over 3.3K additional verifiable, challenging questions from approximately 8K seed samples. Models trained with our synthesized data achieve consistent gains across five out-of-domain visual math reasoning benchmarks, with a significant improvement over baseline models trained on seed data alone. Notably, detailed analysis reveals that the gains are more pronounced on the most challenging evaluation samples, highlighting SynthRL's effectiveness in eliciting deeper and more complex reasoning patterns.