Hi! I'm Niels from the community science team at Hugging Face. I'm opening this PR to improve the documentation for the `linear-bench-mini` dataset.
Specifically, this PR:
- Adds the `text-generation` task category to the metadata.
- Links the research paper: [Agent-Diff: Benchmarking LLM Agents on Enterprise API Tasks via Code Execution with State-Diff-Based Evaluation](https://huggingface.co/papers/2602.11224).
- Adds links to the project website and GitHub repository.
- Provides a sample usage code snippet based on the `agent-diff` Python SDK found in the official repository.
