Scaling Up LLM Reviews for Google Ads Content Moderation

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Qiao, Wei, Dogra, Tushar, Stretcu, Otilia, Lyu, Yu-Han, Fang, Tiantian, Kwon, Dongjin, Lu, Chun-Ta, Luo, Enming, Wang, Yuan, Chia, Chih-Chun, Fuxman, Ariel, Wang, Fangzhou, Krishna, Ranjay, Tek, Mehmet
Natura: Preprint
Pubblicazione: 2024
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866929252955324416
author Qiao, Wei
Dogra, Tushar
Stretcu, Otilia
Lyu, Yu-Han
Fang, Tiantian
Kwon, Dongjin
Lu, Chun-Ta
Luo, Enming
Wang, Yuan
Chia, Chih-Chun
Fuxman, Ariel
Wang, Fangzhou
Krishna, Ranjay
Tek, Mehmet
author_facet Qiao, Wei
Dogra, Tushar
Stretcu, Otilia
Lyu, Yu-Han
Fang, Tiantian
Kwon, Dongjin
Lu, Chun-Ta
Luo, Enming
Wang, Yuan
Chia, Chih-Chun
Fuxman, Ariel
Wang, Fangzhou
Krishna, Ranjay
Tek, Mehmet
contents Large language models (LLMs) are powerful tools for content moderation, but their inference costs and latency make them prohibitive for casual use on large datasets, such as the Google Ads repository. This study proposes a method for scaling up LLM reviews for content moderation in Google Ads. First, we use heuristics to select candidates via filtering and duplicate removal, and create clusters of ads for which we select one representative ad per cluster. We then use LLMs to review only the representative ads. Finally, we propagate the LLM decisions for the representative ads back to their clusters. This method reduces the number of reviews by more than 3 orders of magnitude while achieving a 2x recall compared to a baseline non-LLM model. The success of this approach is a strong function of the representations used in clustering and label propagation; we found that cross-modal similarity representations yield better results than uni-modal representations.
format Preprint
id arxiv_https___arxiv_org_abs_2402_14590
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Scaling Up LLM Reviews for Google Ads Content Moderation
Qiao, Wei
Dogra, Tushar
Stretcu, Otilia
Lyu, Yu-Han
Fang, Tiantian
Kwon, Dongjin
Lu, Chun-Ta
Luo, Enming
Wang, Yuan
Chia, Chih-Chun
Fuxman, Ariel
Wang, Fangzhou
Krishna, Ranjay
Tek, Mehmet
Information Retrieval
Computation and Language
Machine Learning
Large language models (LLMs) are powerful tools for content moderation, but their inference costs and latency make them prohibitive for casual use on large datasets, such as the Google Ads repository. This study proposes a method for scaling up LLM reviews for content moderation in Google Ads. First, we use heuristics to select candidates via filtering and duplicate removal, and create clusters of ads for which we select one representative ad per cluster. We then use LLMs to review only the representative ads. Finally, we propagate the LLM decisions for the representative ads back to their clusters. This method reduces the number of reviews by more than 3 orders of magnitude while achieving a 2x recall compared to a baseline non-LLM model. The success of this approach is a strong function of the representations used in clustering and label propagation; we found that cross-modal similarity representations yield better results than uni-modal representations.
title Scaling Up LLM Reviews for Google Ads Content Moderation
topic Information Retrieval
Computation and Language
Machine Learning
url https://arxiv.org/abs/2402.14590