--- dataset_info: - config_name: finance features: - name: domain dtype: string - name: post_id dtype: string - name: history dtype: string - name: title dtype: string - name: c_root_id_A dtype: string - name: c_root_id_B dtype: string - name: created_at_utc_A dtype: float64 - name: created_at_utc_B dtype: float64 - name: score_A dtype: int64 - name: score_B dtype: int64 - name: human_ref_A dtype: string - name: human_ref_B dtype: string - name: labels dtype: int64 - name: overlap_ratio dtype: float64 - name: seconds_difference dtype: float64 - name: score_ratio dtype: float64 - name: len_ratio dtype: float64 splits: - name: validation_pref num_bytes: 106745033 num_examples: 55518 - name: test_pref num_bytes: 109417097 num_examples: 58603 - name: train_pref num_bytes: 4167127747 num_examples: 2245068 download_size: 323727636 dataset_size: 4383289877 - config_name: finance_test_prompts features: - name: rank dtype: int64 - name: only_comment dtype: bool - name: domain dtype: string - name: post_id dtype: string - name: title dtype: string - name: history dtype: string - name: c_root_id dtype: string - name: created_at_utc dtype: float64 - name: score dtype: int64 - name: human_ref dtype: string splits: - name: test_prompts num_bytes: 6843041 num_examples: 4929 download_size: 3194622 dataset_size: 6843041 - config_name: gender_sexuality features: - name: domain dtype: string - name: post_id dtype: string - name: history dtype: string - name: title dtype: string - name: c_root_id_A dtype: string - name: c_root_id_B dtype: string - name: created_at_utc_A dtype: float64 - name: created_at_utc_B dtype: float64 - name: score_A dtype: int64 - name: score_B dtype: int64 - name: human_ref_A dtype: string - name: human_ref_B dtype: string - name: labels dtype: int64 - name: overlap_ratio dtype: float64 - name: seconds_difference dtype: float64 - name: score_ratio dtype: float64 - name: len_ratio dtype: float64 splits: - name: validation_pref num_bytes: 344495496 num_examples: 200193 - name: test_pref num_bytes: 314580732 num_examples: 182706 - name: train_pref num_bytes: 12750359301 num_examples: 7302867 download_size: 1072838873 dataset_size: 13409435529 - config_name: gender_sexuality_test_prompts features: - name: rank dtype: int64 - name: only_comment dtype: bool - name: domain dtype: string - name: post_id dtype: string - name: title dtype: string - name: history dtype: string - name: c_root_id dtype: string - name: created_at_utc dtype: float64 - name: score dtype: int64 - name: human_ref dtype: string splits: - name: test_prompts num_bytes: 24889982 num_examples: 17485 download_size: 11276178 dataset_size: 24889982 - config_name: history features: - name: domain dtype: string - name: post_id dtype: string - name: history dtype: string - name: title dtype: string - name: c_root_id_A dtype: string - name: c_root_id_B dtype: string - name: created_at_utc_A dtype: float64 - name: created_at_utc_B dtype: float64 - name: score_A dtype: int64 - name: score_B dtype: int64 - name: human_ref_A dtype: string - name: human_ref_B dtype: string - name: labels dtype: int64 - name: overlap_ratio dtype: float64 - name: seconds_difference dtype: float64 - name: score_ratio dtype: float64 - name: len_ratio dtype: float64 splits: - name: validation_pref num_bytes: 11033885 num_examples: 7255 - name: test_pref num_bytes: 8347339 num_examples: 5511 - name: train_pref num_bytes: 470504584 num_examples: 290723 download_size: 48258052 dataset_size: 489885808 - config_name: history_test_prompts features: - name: rank dtype: int64 - name: only_comment dtype: bool - name: domain dtype: string - name: post_id dtype: string - name: title dtype: string - name: history dtype: string - name: c_root_id dtype: string - name: created_at_utc dtype: float64 - name: score dtype: int64 - name: human_ref dtype: string splits: - name: test_prompts num_bytes: 1461426 num_examples: 1027 download_size: 818158 dataset_size: 1461426 - config_name: politics features: - name: domain dtype: string - name: post_id dtype: string - name: history dtype: string - name: title dtype: string - name: c_root_id_A dtype: string - name: c_root_id_B dtype: string - name: created_at_utc_A dtype: float64 - name: created_at_utc_B dtype: float64 - name: score_A dtype: int64 - name: score_B dtype: int64 - name: human_ref_A dtype: string - name: human_ref_B dtype: string - name: labels dtype: int64 - name: overlap_ratio dtype: float64 - name: seconds_difference dtype: float64 - name: score_ratio dtype: float64 - name: len_ratio dtype: float64 splits: - name: validation_pref num_bytes: 144623174 num_examples: 83290 - name: test_pref num_bytes: 162259697 num_examples: 95042 - name: train_pref num_bytes: 5691098839 num_examples: 3411735 download_size: 453464096 dataset_size: 5997981710 - config_name: politics_test_prompts features: - name: rank dtype: int64 - name: only_comment dtype: bool - name: domain dtype: string - name: post_id dtype: string - name: title dtype: string - name: history dtype: string - name: c_root_id dtype: string - name: created_at_utc dtype: float64 - name: score dtype: int64 - name: human_ref dtype: string splits: - name: test_prompts num_bytes: 9753539 num_examples: 8156 download_size: 4811733 dataset_size: 9753539 - config_name: science features: - name: domain dtype: string - name: post_id dtype: string - name: history dtype: string - name: title dtype: string - name: c_root_id_A dtype: string - name: c_root_id_B dtype: string - name: created_at_utc_A dtype: float64 - name: created_at_utc_B dtype: float64 - name: score_A dtype: int64 - name: score_B dtype: int64 - name: human_ref_A dtype: string - name: human_ref_B dtype: string - name: labels dtype: int64 - name: overlap_ratio dtype: float64 - name: seconds_difference dtype: float64 - name: score_ratio dtype: float64 - name: len_ratio dtype: float64 splits: - name: validation_pref num_bytes: 49143432 num_examples: 28225 - name: test_pref num_bytes: 49485277 num_examples: 28595 - name: train_pref num_bytes: 1886530811 num_examples: 1092551 download_size: 214600434 dataset_size: 1985159520 - config_name: science_test_prompts features: - name: rank dtype: int64 - name: only_comment dtype: bool - name: domain dtype: string - name: post_id dtype: string - name: title dtype: string - name: history dtype: string - name: c_root_id dtype: string - name: created_at_utc dtype: float64 - name: score dtype: int64 - name: human_ref dtype: string splits: - name: test_prompts num_bytes: 6695945 num_examples: 5495 download_size: 3505358 dataset_size: 6695945 configs: - config_name: finance data_files: - split: validation_pref path: finance/validation_pref-* - split: test_pref path: finance/test_pref-* - split: train_pref path: finance/train_pref-* - config_name: finance_test_prompts data_files: - split: test_prompts path: finance_test_prompts/test_prompts-* - config_name: gender_sexuality data_files: - split: validation_pref path: gender_sexuality/validation_pref-* - split: test_pref path: gender_sexuality/test_pref-* - split: train_pref path: gender_sexuality/train_pref-* - config_name: gender_sexuality_test_prompts data_files: - split: test_prompts path: gender_sexuality_test_prompts/test_prompts-* - config_name: history data_files: - split: validation_pref path: history/validation_pref-* - split: test_pref path: history/test_pref-* - split: train_pref path: history/train_pref-* - config_name: history_test_prompts data_files: - split: test_prompts path: history_test_prompts/test_prompts-* - config_name: politics data_files: - split: validation_pref path: politics/validation_pref-* - split: test_pref path: politics/test_pref-* - split: train_pref path: politics/train_pref-* - config_name: politics_test_prompts data_files: - split: test_prompts path: politics_test_prompts/test_prompts-* - config_name: science data_files: - split: validation_pref path: science/validation_pref-* - split: test_pref path: science/test_pref-* - split: train_pref path: science/train_pref-* - config_name: science_test_prompts data_files: - split: test_prompts path: science_test_prompts/test_prompts-* license: odc-by language: - en pretty_name: c --- # Dataset Card for "compred" ## Dataset Details ### Dataset Description Conventional algorithms for training language models (LMs) with human feedback rely on preferences that are assumed to account for an average user, disregarding subjectivity and finer-grained variations. Recent studies have raised concerns about aggregating such diverse, often contradictory human feedback to train a single universal reward model, questioning which values or voices the models align with. Finetuning models to maximize such reward results in generic models that produce outputs not preferred by many user groups, as they tend to average out styles and norms. To study this issue, we collect and release **ComPRed**, a question-answering dataset with community-level preferences from Reddit. This dataset facilitates studying diversity in preferences without incurring privacy concerns associated with individual feedback. - **Curated by:** Allen Institute for AI, The Ohio State University, Carnegie Mellon University - **Paper:** [ArXiv](#) - **Repository:** [https://github.com/allenai/compred](https://github.com/allenai/compred) - **Language(s) (NLP):** English - **License:** https://opendatacommons.org/licenses/by/odc_by_1.0_public_text.txt - **Point of Contact:** [Sachin Kumar & Chan Young Park](mailto:kumar.1145@osu.edu, chanyoun@andrew.cmu.edu) ## Uses ComPreD contains five subsets divided based on factors driving diverging user preferences (we followed a similar process as [SHP](https://huggingface.co/datasets/stanfordnlp/SHP) to create this dataset). | Subset(s) | Factor | | -------- | ------- | | politics | Ideologies | | gender_and_sexuality | Demographics | | finance, history | Community Norms | | science | Level of expertise / Community Norms | ### Loading ```python from datasets import load_dataset # load finance train set finance_train_pref = load_dataset("allenai/compred", "finance", split="train_pref") # load finance test prompts finance_test_prompts = load_dataset("allenai/compred", "finance_test_prompts", split="test_prompts") ``` ### Dataset Structure Coming soon ### Data Creation Please refer to our [paper](#) for details on our dataset collection. ## Licensing Information ComPreD is made available under the ODC-BY requiring the user to follow the licenses of the subsequent parts. ## Citation ``` @article{kumar-park2024, title={{Personalized LMs: Aligning Language Models with Diverse Human Preferences}}, author={Sachin Kumar, Chan Young Park, Yulia Tsvetkov, Noah A. Smith, Hannaneh Hajishirzi}, journal={}, year={2024} } ```