AliceMind

Why we included this project

AliceMind is a single place to find the pre-trained models Alibaba's MinD lab has published across NLP, vision-language, and multimodal work. Each model, from StructBERT and VECO to the mPLUG vision-language series, ships with training and inference code plus released checkpoints, so you can reproduce results without chasing separate papers. Researchers comparing encoder-decoder designs get per-model READMEs and evaluation setups that beat working from a bare paper, while teams building Chinese or cross-lingual systems can use these as baselines and fine-tune on their own data. The collection is sprawling and parts of it have aged since publication, so it pays to check individual model folders for how actively each one is maintained.