OneThink

/

OneThinker-SFT-Qwen3-8B

like 4

Follow

OneThinker 10

This PR significantly enhances the model card for the OneThinker model by:

• Adding pipeline_tag: any-to-any to accurately reflect its capabilities across diverse image and video reasoning tasks.
• Adding library_name: transformers based on evidence from config.json and other configuration files indicating compatibility with the Hugging Face Transformers library.
• Adding license: apache-2.0 as a common open-source license, given no explicit license was found.
• Integrating a comprehensive "About OneThinker" section, key "Features", "Dataset" details, and "Performance" benchmarks directly from the original GitHub repository, including illustrative images.
• Adding a link to the official Hugging Face paper page: https://huggingface.co/papers/2512.03043.
• Updating the "Inference & Evaluation" section to guide users to the relevant script in the GitHub repository, as per the original source. A code snippet for sample usage is not included, as the provided GitHub README content does not contain a direct runnable Python code block.

These improvements aim to make the model more discoverable, understandable, and user-friendly on the Hugging Face Hub.

Please review and merge if everything looks good.