TRON: Targeted Rule-Verifiable Online Environments for Visual Reasoning RL
For researchers in visual reasoning and RL, TRON provides a scalable, verifiable training signal that overcomes the budget constraints of static datasets, with demonstrated improvements across multiple models and benchmarks.
TRON introduces an online environment substrate for visual reasoning RL that generates unbounded, verifiable training instances on demand, enabling curriculum-based training. RL post-training with TRON consistently improves performance on ten external multimodal reasoning benchmarks across three model families.
Reinforcement learning (RL) for visual reasoning needs scalable, verifiable, and controllable training signals. Existing visual RL post-training trains on static curated datasets, with fixed image-question-answer samples bounded by their collection budget. In this work, we introduce TRON (Targeted, Rule-verifiable Online eNvironments), an online environment substrate: a training rollout is generated on demand by a controllable generator-verifier program that samples a fresh latent visual state, renders an image, asks a question, and exactly verifies the answer. A single run can therefore draw an unbounded stream of fresh instances at the difficulty level required by the current curriculum. The current TRON suite contains 520 environments organized into five ability buckets (spatial, mathematical, diagram, pattern/logic, and counting); the same substrate supports both a single full model trained on all buckets and per-bucket ability-specialist models, with no additional data collection. We also introduce a substrate analysis covering generation reliability, instance and level diversity, cross-environment near-duplicates, and base-model pass rate by difficulty level. RL post-training with METHOD consistently improves performance on ten external multimodal reasoning benchmarks across Qwen3-VL-4B, Qwen2.5-VL-7B, and MiMo-VL-7B-SFT.