Innovator-VL-8B-Instruct is a multimodal instruction-following large language model designed for scientific understanding and reasoning. The model integrates strong general-purpose vision-language capabilities with enhanced scientific multimodal alignment, while maintaining a fully transparent and reproducible training pipeline. Unlike approaches that rely on large-scale domain-specific pretraining, Innovator-VL-8B-Instruct achieves competitive scientific performance using high-quality instruction tuning, without additional scientific text continued pretraining. — Multimodal Alignment: LLaVA-1.5 (558K) — Mid-training: LLaVA-OneVision-1.5 (85M) — Instruction Tuning: High-quality multimodal and scientific instruction data (~46M) No additional scientific text continued pretraining is applied. Below is a minimal example to run multimodal inference (image + text) with transformers. «`bibtex @article{wen2026innovator, title={Innovator-VL: A Multimodal Large Language Model for Scientific Discovery}, author={Wen, Zichen and Yang, Boxue and Chen, Shuang and Zhang, Yaojie and Han, Yuhang and Ke, Junlong and Wang, Cong and others}, journal={arXiv preprint arXiv:2601.19325}, year={2026} }
Модальности:
Генерация текста
Области применения:
Диалог / чат Следование инструкциям
Задача: Генерация текста
Автор: InnovatorLab
Теги: conversational, custom_code, en, zh
Лайков: 4 | Загрузок: 17
Описание основано на материалах HuggingFace. Перевод выполнен автоматически.
