This PR adds a comprehensive model card for the LEGO model, presented in the paper [Advancing 3D Scene Understanding with MV-ScanQA Multi-View Reasoning Evaluation and TripAlign Pre-training Dataset](https://huggingface.co/papers/2508.11058).
It includes:
- The appropriate `pipeline_tag`: `image-text-to-text`, allowing users to find the model easily on the Hub.
- The `library_name`: `transformers`, indicating compatibility with the 🤗 Transformers library.
- The `license`: `CC-BY-4.0` as specified in the repository.
- Links to the paper, project page, and the official GitHub repository for more detailed information and code.
- An overview of the model and its capabilities based on the paper abstract and GitHub README.
- A sample Python usage example demonstrating how to load and use the LoRA adapter with its Fuyu base model.
Please review and merge this PR if everything looks good.
update reproduction pre-trained checkpoints from cleaned script
