Add paper link, GitHub link, and task category metadata

Hi! I'm Niels from the Hugging Face community science team.

This pull request improves the dataset card by:
1. Adding the `image-text-to-text` task category to the metadata.
2. Linking the dataset to the official paper: "Spec-o3: A Tool-Augmented Vision-Language Agent for Rare Celestial Object Candidate Vetting via Automated Spectral Inspection".
3. Adding a link to the GitHub repository for the project.

These changes help users discover the dataset via the Hub's task filters and provide the necessary context for the research.

README.md

@@ -1,3 +1,17 @@
+---
+language:
+- en
+task_categories:
+- image-text-to-text
+tags:
+- astronomy
+- science
+---
+
+# SpecVI-Bench
+
+[Paper](https://huggingface.co/papers/2601.06498) | [Code](https://github.com/Maxwell-Jia/spec-o3)
+
 ## Summary
 This dataset contains spectral classification tasks built for tool-using models.
 Each example includes a full-spectrum image, a chat-style prompt, and ground-truth
@@ -53,4 +67,4 @@ Each example contains the following fields:
 
 ## Notes
 - `images` contains the full-spectrum overview image for the example.
-- `extra_info` provides raw arrays to recreate spectrum plots.
+- `extra_info` provides raw arrays to recreate spectrum plots.