--- dataset_info: - config_name: ACSEmployment features: - name: id dtype: int64 - name: description dtype: string - name: instruction dtype: string - name: question dtype: string - name: choices sequence: string length: 2 - name: answer dtype: string - name: answer_key dtype: string - name: choice_question_prompt dtype: string - name: numeric_question dtype: string - name: label dtype: int64 - name: numeric_question_prompt dtype: string - name: AGEP dtype: int64 - name: SCHL dtype: float64 - name: MAR dtype: int64 - name: RELP dtype: int64 - name: DIS dtype: int64 - name: ESP dtype: float64 - name: CIT dtype: int64 - name: MIG dtype: float64 - name: MIL dtype: float64 - name: ANC dtype: int64 - name: NATIVITY dtype: int64 - name: DEAR dtype: int64 - name: DEYE dtype: int64 - name: DREM dtype: float64 - name: SEX dtype: int64 - name: RAC1P dtype: int64 splits: - name: train num_bytes: 4216821661 num_examples: 2588885 - name: validation num_bytes: 527108317 num_examples: 323611 - name: test num_bytes: 527106993 num_examples: 323611 download_size: 326506898 dataset_size: 5271036971 - config_name: ACSIncome features: - name: id dtype: int64 - name: description dtype: string - name: instruction dtype: string - name: question dtype: string - name: choices sequence: string length: 2 - name: answer dtype: string - name: answer_key dtype: string - name: choice_question_prompt dtype: string - name: numeric_question dtype: string - name: label dtype: int64 - name: numeric_question_prompt dtype: string - name: AGEP dtype: int64 - name: COW dtype: int64 - name: SCHL dtype: int64 - name: MAR dtype: int64 - name: OCCP dtype: int64 - name: POBP dtype: int64 - name: RELP dtype: int64 - name: WKHP dtype: int64 - name: SEX dtype: int64 - name: RAC1P dtype: int64 splits: - name: train num_bytes: 1667556978 num_examples: 1331600 - name: validation num_bytes: 208460485 num_examples: 166450 - name: test num_bytes: 208446084 num_examples: 166450 download_size: 155811003 dataset_size: 2084463547 - config_name: ACSMobility features: - name: id dtype: int64 - name: description dtype: string - name: instruction dtype: string - name: question dtype: string - name: choices sequence: string length: 2 - name: answer dtype: string - name: answer_key dtype: string - name: choice_question_prompt dtype: string - name: numeric_question dtype: string - name: label dtype: int64 - name: numeric_question_prompt dtype: string - name: AGEP dtype: int64 - name: SCHL dtype: float64 - name: MAR dtype: int64 - name: SEX dtype: int64 - name: DIS dtype: int64 - name: ESP dtype: float64 - name: CIT dtype: int64 - name: MIL dtype: float64 - name: ANC dtype: int64 - name: NATIVITY dtype: int64 - name: RELP dtype: int64 - name: DEAR dtype: int64 - name: DEYE dtype: int64 - name: DREM dtype: float64 - name: RAC1P dtype: int64 - name: GCL dtype: float64 - name: COW dtype: float64 - name: ESR dtype: float64 - name: WKHP dtype: float64 - name: JWMNP dtype: float64 - name: PINCP dtype: float64 splits: - name: train num_bytes: 1028069699 num_examples: 496749 - name: validation num_bytes: 128514698 num_examples: 62094 - name: test num_bytes: 128509092 num_examples: 62094 download_size: 86257467 dataset_size: 1285093489 - config_name: ACSPublicCoverage features: - name: id dtype: int64 - name: description dtype: string - name: instruction dtype: string - name: question dtype: string - name: choices sequence: string length: 2 - name: answer dtype: string - name: answer_key dtype: string - name: choice_question_prompt dtype: string - name: numeric_question dtype: string - name: label dtype: int64 - name: numeric_question_prompt dtype: string - name: AGEP dtype: int64 - name: SCHL dtype: float64 - name: MAR dtype: int64 - name: SEX dtype: int64 - name: DIS dtype: int64 - name: ESP dtype: float64 - name: CIT dtype: int64 - name: MIG dtype: float64 - name: MIL dtype: float64 - name: ANC dtype: int64 - name: NATIVITY dtype: int64 - name: DEAR dtype: int64 - name: DEYE dtype: int64 - name: DREM dtype: float64 - name: PINCP dtype: float64 - name: ESR dtype: float64 - name: ST dtype: int64 - name: FER dtype: float64 - name: RAC1P dtype: int64 splits: - name: train num_bytes: 1775349595 num_examples: 910631 - name: validation num_bytes: 221922965 num_examples: 113829 - name: test num_bytes: 221932893 num_examples: 113829 download_size: 138771081 dataset_size: 2219205453 - config_name: ACSTravelTime features: - name: id dtype: int64 - name: description dtype: string - name: instruction dtype: string - name: question dtype: string - name: choices sequence: string length: 2 - name: answer dtype: string - name: answer_key dtype: string - name: choice_question_prompt dtype: string - name: numeric_question dtype: string - name: label dtype: int64 - name: numeric_question_prompt dtype: string - name: AGEP dtype: int64 - name: SCHL dtype: float64 - name: MAR dtype: int64 - name: SEX dtype: int64 - name: DIS dtype: int64 - name: ESP dtype: float64 - name: MIG dtype: float64 - name: RELP dtype: int64 - name: RAC1P dtype: int64 - name: PUMA dtype: int64 - name: ST dtype: int64 - name: CIT dtype: int64 - name: OCCP dtype: float64 - name: JWTR dtype: float64 - name: POWPUMA dtype: float64 - name: POVPIP dtype: float64 splits: - name: train num_bytes: 2039450205 num_examples: 1173318 - name: validation num_bytes: 254946722 num_examples: 146665 - name: test num_bytes: 254948776 num_examples: 146665 download_size: 215781517 dataset_size: 2549345703 configs: - config_name: ACSEmployment data_files: - split: train path: ACSEmployment/train-* - split: validation path: ACSEmployment/validation-* - split: test path: ACSEmployment/test-* - config_name: ACSIncome data_files: - split: train path: ACSIncome/train-* - split: validation path: ACSIncome/validation-* - split: test path: ACSIncome/test-* default: true - config_name: ACSMobility data_files: - split: train path: ACSMobility/train-* - split: validation path: ACSMobility/validation-* - split: test path: ACSMobility/test-* - config_name: ACSPublicCoverage data_files: - split: train path: ACSPublicCoverage/train-* - split: validation path: ACSPublicCoverage/validation-* - split: test path: ACSPublicCoverage/test-* - config_name: ACSTravelTime data_files: - split: train path: ACSTravelTime/train-* - split: validation path: ACSTravelTime/validation-* - split: test path: ACSTravelTime/test-* license: mit task_categories: - question-answering - text-classification - feature-extraction - zero-shot-classification language: - en pretty_name: Folktexts real-world unrealizable classification tasks size_categories: - 1M [Folktexts](https://github.com/socialfoundations/folktexts) is a suite of Q&A datasets with natural outcome uncertainty, aimed at evaluating LLMs' calibration on unrealizable tasks. The *folktexts* datasets are derived from US Census data products. Namely, the datasets made available here are derived from the [2018 Public Use Microdata Sample](https://www.census.gov/programs-surveys/acs/microdata/documentation/2018.html) (PUMS). Individual features are mapped to natural text using the respective [codebook](https://www2.census.gov/programs-surveys/acs/tech_docs/pums/data_dict/PUMS_Data_Dictionary_2018.pdf). Each task relates to predicting different individual characteristics (e.g., income, employment) from a set of demographic features (e.g., age, race, education, occupation). Importantly, every task has natural outcome uncertainty. That is, in general, the features describing each row do not uniquely determine the task's label. For calibrated models to perform well on this task, the model must correctly output nuanced scores between 0 and 1, instead of simply outputting discrete labels 0 or 1. Namely, we make available the following tasks in natural language Q&A format: - `ACSIncome`: Predict whether a working adult earns above $50,000 yearly. - `ACSEmployment`: Predict whether an adult is an employed civilian. - `ACSPublicCoverage`: Predict individual public health insurance coverage. - `ACSMobility`: Predict whether an individual changed address within the last year. - `ACSTravelTime`: Predict whether an employed adult has a work commute time longer than 20 minutes. These tasks follow the same naming and feature/target columns as the [folktables](https://github.com/socialfoundations/folktables) tabular datasets proposed by [Ding et al. (2021)](https://proceedings.neurips.cc/paper_files/paper/2021/file/32e54441e6382a7fbacbbbaf3c450059-Paper.pdf). The folktables tabular datasets have seen prevalent use in the algorithmic fairness and distribution shift communities. We make available natural language Q&A versions of these tasks. The datasets are made available in standard multiple-choice Q&A format (columns `question`, `choices`, `answer`, `answer_key`, and `choice_question_prompt`), as well as in numeric Q&A format (columns `numeric_question`, `numeric_question_prompt`, and `label`). The numeric prompting (also known as *verbalized prompting*) is known to improve calibration of zero-shot LLM risk scores [[Tian et al., EMNLP 2023](https://openreview.net/forum?id=g3faCfrwm7); [Cruz et al., NeurIPS 2024](https://arxiv.org/pdf/2407.14614)]. **The accompanying [`folktexts` python package](https://github.com/socialfoundations/folktexts) eases customization, evaluation, and benchmarking with these datasets.** Table of contents: - [Dataset Details](#dataset-details) - [Uses](#uses) - [Dataset Structure](#dataset-structure) - [Dataset Creation](#dataset-creation) - [Citation](#citation) - [More Information](#more-information) ## Dataset Details ### Dataset Description - **Language(s) (NLP):** English - **License:** Code is licensed under the [MIT license](https://github.com/socialfoundations/folktexts/blob/main/LICENSE); Data license is governed by the U.S. Census Bureau [terms of service](https://www.census.gov/data/developers/about/terms-of-service.html). ### Dataset Sources - **Repository:** https://github.com/socialfoundations/folktexts - **Paper:** https://arxiv.org/pdf/2407.14614 - **Data source:** [2018 American Community Survey Public Use Microdata Sample](https://www.census.gov/programs-surveys/acs/microdata/documentation/2018.html) ## Uses The datasets were originally used to evaluate LLMs' ability to produce calibrated and accurate risk scores in the [Cruz et al. (2024)](https://arxiv.org/pdf/2407.14614) paper. Other potential uses include evaluating the fairness of LLMs' decisions, as individual rows feature protected demographic attributes such as `sex` and `race`. ## Dataset Structure **Description of dataset columns:** - `id`: A unique row identifier. - `description`: A textual description of an individual's features, following a bulleted-list format. - `instruction`: The instruction used for zero-shot LLM prompting (should be pre-appended to the row description). - `question`: A question relating to the task's target column. - `choices`: A list of two answer options relating to the above question. - `answer`: The correct answer from the above list of answer options. - `answer_key`: The correct answer key; i.e., `A` for the first choice, or `B` for the second choice. - `choice_question_prompt`: The full multiple-choice Q&A text string used for LLM prompting. - `numeric_question`: A version of the question that prompts for a *numeric output* instead of a *discrete choice output*. - `label`: The task's label. This is the correct output to the above numeric question. - `numeric_question_prompt`: The full numeric Q&A text string used for LLM prompting. - ``: All other columns correspond to the tabular features in this task. Each of these features will also appear in text form on the above description column. The dataset was randomly split in `training`, `test`, and `validation` data, following an 80%/10%/10% split. Only the `test` split should be used to evaluate zero-shot LLM performance. The `training` split can be used for fine-tuning, or for fitting traditional supervised ML models on the tabular columns for metric baselines. The `validation` split should be used for hyperparameter tuning, feature engineering or any other model improvement loop. ## Dataset Creation ### Source Data The datasets are based on publicly available data from the American Community Survey (ACS) Public Use Microdata Sample (PUMS), namely the [2018 ACS 1-year PUMS files](https://www.census.gov/programs-surveys/acs/microdata/documentation.2018.html#list-tab-1370939201). #### Data Collection and Processing The categorical values were mapped to meaningful natural language representations using the `folktexts` package, which in turn uses the official [ACS PUMS codebook](https://www2.census.gov/programs-surveys/acs/tech_docs/pums/data_dict/PUMS_Data_Dictionary_2018.pdf). The data download and processing was aided by the `folktables` python package, which in turn uses the official US Census web API. #### Who are the source data producers? U.S. Census Bureau. ## Citation If you find this useful in your research, please consider citing the following paper: ```bib @inproceedings{ cruz2024evaluating, title={Evaluating language models as risk scores}, author={Andr{\'e} F Cruz and Moritz Hardt and Celestine Mendler-D{\"u}nner}, booktitle={The Thirty-eight Conference on Neural Information Processing Systems Datasets and Benchmarks Track}, year={2024}, url={https://openreview.net/forum?id=qrZxL3Bto9} } ``` ## More Information More information is available in the [`folktexts`](https://github.com/socialfoundations/folktexts) package repository and the [accompanying paper](https://arxiv.org/pdf/2407.14614). ### Dataset Card Authors [André F. Cruz](https://github.com/andrefcruz)