Post-processing for Individual Fairness

Authors: Felix Petersen, Debarghya Mukherjee, Yuekai Sun, Mikhail Yurochkin

NeurIPS 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Empirically, our post-processing algorithms correct individual biases in large-scale NLP models such as BERT, while preserving accuracy.
Researcher Affiliation Collaboration Felix Petersen Debarghya Mukherjee University of Konstanz University of Michigan felix.petersen@uni.kn mdeb@umich.edu Yuekai Sun Mikhail Yurochkin University of Michigan IBM Research, MIT-IBM Watson AI Lab yuekai@umich.edu mikhail.yurochkin@ibm.com
Pseudocode No The paper describes algorithmic steps and solutions in text and equations but does not include a structured pseudocode or algorithm block.
Open Source Code Yes The implementation of this work is available at github.com/Felix-Petersen/fairness-post-processing.
Open Datasets Yes We replicate the experiments of Yurochkin et al. [12] on Bios [40] and Toxicity1 data sets.
Dataset Splits No The paper mentions 'validation data' for hyperparameter selection but does not provide specific details on the dataset splits (e.g., percentages or counts) for training, validation, and testing.
Hardware Specification No The paper mentions general computational resources but does not specify the hardware used for running the experiments (e.g., exact GPU/CPU models, memory, or cloud instances).
Software Dependencies No The paper mentions software like CVXPY and GloVe embeddings but does not provide specific version numbers for any software dependencies used in the experiments.
Experiment Setup Yes We evaluate the fairness-accuracy trade-offs for a range of threshold parameters τ (for GLIF and GLIF-NRW) and for a range of Lipschitz-constants L (for IF-constraints) in Figure 2.