Innovator-VL: A Multimodal Large Language Model for Scientific Discovery

TL;DR

Innovator-VL is a transparent, data-efficient multimodal large language model that advances scientific reasoning across domains without relying on massive pretraining, emphasizing principled training and reproducibility.

Contribution

The paper introduces Innovator-VL, a scientific multimodal LLM with a fully transparent training pipeline and high performance achieved with limited data, challenging the need for large-scale pretraining.

Findings

Achieves competitive scientific reasoning with fewer than five million samples.

Demonstrates strong generalization across scientific and vision tasks.

Provides a reproducible training pipeline for community extension.

Abstract

We present Innovator-VL, a scientific multimodal large language model designed to advance understanding and reasoning across diverse scientific domains while maintaining excellent performance on general vision tasks. Contrary to the trend of relying on massive domain-specific pretraining and opaque pipelines, our work demonstrates that principled training design and transparent methodology can yield strong scientific intelligence with substantially reduced data requirements. (i) First, we provide a fully transparent, end-to-end reproducible training pipeline, covering data collection, cleaning, preprocessing, supervised fine-tuning, reinforcement learning, and evaluation, along with detailed optimization recipes. This facilitates systematic extension by the community. (ii) Second, Innovator-VL exhibits remarkable data efficiency, achieving competitive performance on various scientific tasks using fewer than five million curated samples without large-scale pretraining. These results highlight that effective reasoning can be achieved through principled data selection rather than indiscriminate scaling. (iii) Third, Innovator-VL demonstrates strong generalization, achieving competitive performance on general vision, multimodal reasoning, and scientific benchmarks. This indicates that scientific alignment can be integrated into a unified model without compromising general-purpose capabilities. Our practices suggest that efficient, reproducible, and high-performing scientific multimodal models can be built even without large-scale data, providing a practical foundation for future research.