Add task categories and link to paper

README.md

@@ -1,24 +1,26 @@
 ---
-license: mit
 language:
-  - en
-tags:
-  - machine-learning
-  - academic-papers
-  - slides
-  - multimodal
-  - benchmark
-  - nlp
-  - computer-vision
-  - oral-presentations
-pretty_name: ArcBench ML Conference Oral Paper-Presentation Benchmark
+- en
+license: mit
 size_categories:
-  - n<1K
+- n<1K
+pretty_name: ArcBench ML Conference Oral Paper-Presentation Benchmark
+task_categories:
+- other
+tags:
+- machine-learning
+- academic-papers
+- slides
+- multimodal
+- benchmark
+- nlp
+- computer-vision
+- oral-presentations
 ---
 
 # ArcBench: ML Conference Oral Paper-Presentation Benchmark
 
-This benchmark is from the paper **Narrative-Driven Paper-to-Slide Generation via ArcDeck**.
+This benchmark is from the paper **[Narrative-Driven Paper-to-Slide Generation via ArcDeck](https://huggingface.co/papers/2604.11969)**.
 
 A curated benchmark dataset of **100 oral presentation papers** from top-tier machine learning conferences (CVPR, ICCV, ICLR, ICML, NeurIPS), spanning 2022–2025. Each entry includes the full paper PDF, presentation slides PDF, and rich metadata.
 
@@ -167,4 +169,4 @@ If you use this dataset in your research, please cite:
 ## License
 
 MIT.
-Individual paper and slide PDFs remain under their original authors' copyright. Please consult each paper's license before redistribution.
+Individual paper and slide PDFs remain under their original authors' copyright. Please consult each paper's license before redistribution.