BendVLM: Test-Time Debiasing of Vision-Language Embeddings

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Gerych, Walter, Zhang, Haoran, Hamidieh, Kimia, Pan, Eileen, Sharma, Maanas, Hartvigsen, Thomas, Ghassemi, Marzyeh
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866917829869043712
author Gerych, Walter
Zhang, Haoran
Hamidieh, Kimia
Pan, Eileen
Sharma, Maanas
Hartvigsen, Thomas
Ghassemi, Marzyeh
author_facet Gerych, Walter
Zhang, Haoran
Hamidieh, Kimia
Pan, Eileen
Sharma, Maanas
Hartvigsen, Thomas
Ghassemi, Marzyeh
contents Vision-language model (VLM) embeddings have been shown to encode biases present in their training data, such as societal biases that prescribe negative characteristics to members of various racial and gender identities. VLMs are being quickly adopted for a variety of tasks ranging from few-shot classification to text-guided image generation, making debiasing VLM embeddings crucial. Debiasing approaches that fine-tune the VLM often suffer from catastrophic forgetting. On the other hand, fine-tuning-free methods typically utilize a "one-size-fits-all" approach that assumes that correlation with the spurious attribute can be explained using a single linear direction across all possible inputs. In this work, we propose Bend-VLM, a nonlinear, fine-tuning-free approach for VLM embedding debiasing that tailors the debiasing operation to each unique input. This allows for a more flexible debiasing approach. Additionally, we do not require knowledge of the set of inputs a priori to inference time, making our method more appropriate for online, open-set tasks such as retrieval and text guided image generation.
format Preprint
id arxiv_https___arxiv_org_abs_2411_04420
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle BendVLM: Test-Time Debiasing of Vision-Language Embeddings
Gerych, Walter
Zhang, Haoran
Hamidieh, Kimia
Pan, Eileen
Sharma, Maanas
Hartvigsen, Thomas
Ghassemi, Marzyeh
Computer Vision and Pattern Recognition
Machine Learning
Vision-language model (VLM) embeddings have been shown to encode biases present in their training data, such as societal biases that prescribe negative characteristics to members of various racial and gender identities. VLMs are being quickly adopted for a variety of tasks ranging from few-shot classification to text-guided image generation, making debiasing VLM embeddings crucial. Debiasing approaches that fine-tune the VLM often suffer from catastrophic forgetting. On the other hand, fine-tuning-free methods typically utilize a "one-size-fits-all" approach that assumes that correlation with the spurious attribute can be explained using a single linear direction across all possible inputs. In this work, we propose Bend-VLM, a nonlinear, fine-tuning-free approach for VLM embedding debiasing that tailors the debiasing operation to each unique input. This allows for a more flexible debiasing approach. Additionally, we do not require knowledge of the set of inputs a priori to inference time, making our method more appropriate for online, open-set tasks such as retrieval and text guided image generation.
title BendVLM: Test-Time Debiasing of Vision-Language Embeddings
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2411.04420