Update README.md

README.md

@@ -1,14 +1,22 @@
-    DatasetDict({
-        train_prefs: Dataset({
-            features: ['post_id', 'domain', 'upvote_ratio', 'history', 'c_root_id_A', 'c_root_id_B', 'created_at_utc_A', 'created_at_utc_B', 'score_A', 'score_B', 'human_ref_A', 'human_ref_B', 'labels', 'seconds_difference', 'score_ratio'],
-            num_rows: 80000
-        })
-        test_prefs: Dataset({
-            features: ['post_id', 'domain', 'upvote_ratio', 'history', 'c_root_id_A', 'c_root_id_B', 'created_at_utc_A', 'created_at_utc_B', 'score_A', 'score_B', 'human_ref_A', 'human_ref_B', 'labels', 'seconds_difference', 'score_ratio'],
-            num_rows: 2000
-        })
+# Dataset Card for "shp_filtered_MMPO"
 
+This is the filtered version of SHP dataset, which was used to train MMPO, as introduced in the paper below:
 
-# Dataset Card for "shp_filtered_uniform_large"
+**Margin Matching Preference Optimization: Enhanced Model Alignment with Granular Feedback** <br>
+Kyuyoung Kim*, Ah Jeong Seo*, Hao Liu, Jinwoo Shin, Kimin Lee  <br>
+*In EMNLP 2024 Findings*
 
-[More Information needed](https://github.com/huggingface/datasets/blob/main/CONTRIBUTING.md#how-to-contribute-to-the-dataset-cards)
+
+## Dataset Description
+
+The original [SHP dataset](https://huggingface.co/datasets/stanfordnlp/SHP) consists of 385k collective human preferences over responses to questions/instructions in 18 different subject areas, from cooking to legal advice. 
+To create SHP filtered dataset to train MMPO, we extracted a subset of size 55k, following [Ethayarajh et al. (2022)](https://arxiv.org/abs/2110.08420) and [Sun et al. (2023)](https://arxiv.org/abs/2310.05910).
+
+However, we sampled uniformly across score differences to evaluate the methods over a wide range of quality margins, which is different from prior works that trained only on preferences with significant score differences.
+Upon analyzing the distribution of score differences in the SHP dataset, we found that 50% of the data had relatively small differences. 
+Therefore, to check if the model can be optimized effectively with datasets containing many preferences with low confidence, 
+we employed a method of **sampling proportional to the score difference distribution of the original SHP**.
+
+
+More details can be found in the paper referenced above. 
+Additionally, you can find more filtering details about the SHP dataset in the [official code](https://github.com/kykim0/margin-matching-pref-opt).