Towards trustworthy LLMs: a review on debiasing and dehallucinating in large language models
Z Lin, S Guan, W Zhang, H Zhang, Y Li… - Artificial Intelligence …, 2024 - Springer
Recently, large language models (LLMs) have attracted considerable attention due to their
remarkable capabilities. However, LLMs' generation of biased or hallucinatory content …
remarkable capabilities. However, LLMs' generation of biased or hallucinatory content …
Language (technology) is power: A critical survey of" bias" in nlp
We survey 146 papers analyzing" bias" in NLP systems, finding that their motivations are
often vague, inconsistent, and lacking in normative reasoning, despite the fact that …
often vague, inconsistent, and lacking in normative reasoning, despite the fact that …
Unlearning bias in language models by partitioning gradients
Recent research has shown that large-scale pretrained language models, specifically
transformers, tend to exhibit issues relating to racism, sexism, religion bias, and toxicity in …
transformers, tend to exhibit issues relating to racism, sexism, religion bias, and toxicity in …
A survey on gender bias in natural language processing
Language can be used as a means of reproducing and enforcing harmful stereotypes and
biases and has been analysed as such in numerous research. In this paper, we present a …
biases and has been analysed as such in numerous research. In this paper, we present a …
RedditBias: A real-world resource for bias evaluation and debiasing of conversational language models
Text representation models are prone to exhibit a range of societal biases, reflecting the non-
controlled and biased nature of the underlying pretraining data, which consequently leads to …
controlled and biased nature of the underlying pretraining data, which consequently leads to …
Sustainable modular debiasing of language models
Unfair stereotypical biases (eg, gender, racial, or religious biases) encoded in modern
pretrained language models (PLMs) have negative ethical implications for widespread …
pretrained language models (PLMs) have negative ethical implications for widespread …
A survey of race, racism, and anti-racism in NLP
Despite inextricable ties between race and language, little work has considered race in NLP
research and development. In this work, we survey 79 papers from the ACL anthology that …
research and development. In this work, we survey 79 papers from the ACL anthology that …
Debiasing pre-trained contextualised embeddings
In comparison to the numerous debiasing methods proposed for the static non-
contextualised word embeddings, the discriminative biases in contextualised embeddings …
contextualised word embeddings, the discriminative biases in contextualised embeddings …
[PDF][PDF] Survey on sociodemographic bias in natural language processing
Deep neural networks often learn unintended bias during training, which might have harmful
effects when deployed in realworld settings. This work surveys 214 papers related to …
effects when deployed in realworld settings. This work surveys 214 papers related to …
Unmasking the mask–evaluating social biases in masked language models
Abstract Masked Language Models (MLMs) have shown superior performances in
numerous downstream Natural Language Processing (NLP) tasks. Unfortunately, MLMs …
numerous downstream Natural Language Processing (NLP) tasks. Unfortunately, MLMs …