Datasets:

---

walktaster

/

LTD_Bench

like 1

This PR significantly enriches the dataset card for LTD-Bench by:

• Adding a descriptive introduction to the dataset.
• Linking to the official paper: https://huggingface.co/papers/2511.02347
• Linking to the GitHub repository: https://github.com/walktaster/LTD-Bench
• Including the full paper abstract for comprehensive context.
• Specifying text-generation and image-to-text as task_categories to reflect the benchmark's focus on bidirectional language-spatial mapping and visual outputs.
• Adding relevant tags such as benchmark, llm-evaluation, spatial-reasoning, and multimodal for better discoverability.
• Adding language: en to indicate the primary language.
• Providing a detailed "Sample Usage" section, extracted directly from the GitHub README, which includes steps for environment setup, running model inference, and performing evaluation.

These additions will make the dataset more discoverable and informative for researchers and practitioners on the Hugging Face Hub.