REVIEW 5 cited by
Perturbation Augmentation for Fairer NLP
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
Unwanted and often harmful social biases are becoming ever more salient in NLP research, affecting both models and datasets. In this work, we ask whether training on demographically perturbed data leads to fairer language models. We collect a large dataset of human annotated text perturbations and train a neural perturbation model, which we show outperforms heuristic alternatives. We find that (i) language models (LMs) pre-trained on demographically perturbed corpora are typically more fair, and (ii) LMs finetuned on perturbed GLUE datasets exhibit less demographic bias on downstream tasks, and (iii) fairness improvements do not come at the expense of performance on downstream tasks. Lastly, we discuss outstanding questions about how best to evaluate the (un)fairness of large language models. We hope that this exploration of neural demographic perturbation will help drive more improvement towards fairer NLP.
Forward citations
Cited by 5 Pith papers
-
LLMs on Trial: Evaluating Judicial Fairness for Large Language Models
A new 177,100-case benchmark shows that 16 LLMs systematically vary criminal sentences based on extra-legal demographic and procedural details, revealing pervasive judicial unfairness.
-
Investigating Intersectional Bias in Large Language Models using Confidence Disparities in Coreference Resolution
The abstract claims LLMs show up to 40% coreference confidence disparities across intersectional identities, but the article body is an unrelated paper on robotic fruit handling.
-
Mitigating Confounding in Speech-Based Dementia Detection through Weight Masking
Masking weights that react to gender in a fine-tuned BERT reduces gender gaps in dementia predictions while keeping most of the detection accuracy.
-
LFTF: Locating First and Then Fine-Tuning for Mitigating Gender Bias in Large Language Models
A block-localizing fine-tuning method for gender debiasing is presented, but its stated loss is inconsistent with its reported behavior and the evaluation tables contain duplicate rows.
-
The Societal Impact of Foundation Models: Advancing Evidence-based AI Policy
A dissertation that synthesizes prior work on foundation models into a three-part framework: conceptual framing, empirical measurement (HELM, FMTI), and evidence-based AI policy.
Discussion (0). Continue with ORCID to comment.