Improve dataset card: Add paper, code links, task categories, and description

README.md

@@ -3,35 +3,53 @@ configs:
 - config_name: default
   data_files:
   - split: law
-    path: "law.json"
+    path: law.json
   - split: psychology
-    path: "psychology.json"
+    path: psychology.json
   - split: chemistry
-    path: "chemistry.json"
+    path: chemistry.json
   - split: biology
-    path: "biology.json"
+    path: biology.json
   - split: physics
-    path: "physics.json"
+    path: physics.json
   - split: history
-    path: "history.json"
+    path: history.json
   - split: economics
-    path: "economics.json"
+    path: economics.json
   - split: math
-    path: "math.json"
+    path: math.json
   - split: business
-    path: "business.json"
+    path: business.json
   - split: philosophy
-    path: "philosophy.json"
+    path: philosophy.json
   - split: health
-    path: "health.json"
+    path: health.json
   - split: engineering
-    path: "engineering.json"
+    path: engineering.json
   - split: computer_science
-    path: "computer_science.json"
+    path: computer_science.json
   - split: other
-    path: "other.json"
+    path: other.json
+task_categories:
+- text-ranking
+language:
+- en
+tags:
+- reward-modeling
+- preference-alignment
+- multi-domain
+- llm-evaluation
 ---
-# Reward of [test](https://huggingface.co/datasets/dongboklee/test) split extracted by gPRM-8B: gPRM-8B-test
+
+# gPRM-8B-test: Reward Dataset for Multi-Domain Test-Time Scaling
+
+This dataset contains the reward scores for the "test" split, extracted by the gPRM-8B model, as described in the paper [Rethinking Reward Models for Multi-Domain Test-Time Scaling](https://huggingface.co/papers/2510.00492). This work evaluates various reward model variants across 14 diverse domains to assess the reliability of large language models (LLMs) during test-time scaling.
+
+The `gPRM-8B-test` dataset specifically provides rewards assigned by the gPRM-8B model to Chains of Thought (CoTs) generated by `Llama3.1-8B-Instruct` across multiple domains like law, psychology, chemistry, and more.
+
+**Paper:** [Rethinking Reward Models for Multi-Domain Test-Time Scaling](https://huggingface.co/papers/2510.00492)
+**Code:** [https://github.com/db-Lee/Multi-RM](https://github.com/db-Lee/Multi-RM)
+
 ## Usage
 ```python
 from datasets import load_dataset
@@ -42,3 +60,14 @@ dataset = load_dataset("dongboklee/gPRM-8B-test")
 # Load specific domain
 law_dataset = load_dataset("dongboklee/gPRM-8B-test", split="law")
 ```
+
+## Citation
+If you use this dataset in your research, please cite the original paper:
+```bibtex
+@article{multi-rm,
+  title   = {Rethinking Reward Models for Multi-Domain Test-Time Scaling},
+  author  = {Lee, Dong Bok and Lee, Seanie and Park, Sangwoo and Kang, Minki and Baek, Jinheon and Kim, Dongki and Wagner, Dominik and Jin, Jiongdao and Lee, Heejun and Bocklet, Tobias and Wang, Jinyu and Fu, Jingjing and Hwang, Sung Ju and Bian, Jiang and Song, Lei},
+  journal = {arXiv preprint arXiv:2510.00492},
+  year    = {2025}
+}
+```