Update pipeline tag to `image-text-to-text` and add `robotics` tag
This PR updates the model card for RoboBrain 2.0 to improve its discoverability and accuracy on the Hugging Face Hub.
The pipeline_tag has been updated from robotics to image-text-to-text. This more accurately reflects the model's input modalities (multi-image, long video, text instructions) and its text-based outputs (e.g., answers, structured plans, coordinates) across various tasks such as visual question answering, visual grounding, and embodied reasoning.
Additionally, the robotics tag has been moved from pipeline_tag to the general tags list. This ensures that the model remains discoverable for users interested in robotics and embodied AI, while correctly categorizing its core functionality as a multimodal language model.
The model card content has been updated to reflect the latest information from the project's GitHub README, including a recent news update.