Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Liu, Mingchao, Sun, Yu, Sun, Ruixiao, Dong, Xin, Shen, Xiang, Wang, Hongwei, Xiong, Hongyu, Song, Yang
Format: Preprint
Veröffentlicht: 2024
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2412.15251
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866913083899772928
author Liu, Mingchao
Sun, Yu
Sun, Ruixiao
Dong, Xin
Shen, Xiang
Wang, Hongwei
Xiong, Hongyu
Song, Yang
author_facet Liu, Mingchao
Sun, Yu
Sun, Ruixiao
Dong, Xin
Shen, Xiang
Wang, Hongwei
Xiong, Hongyu
Song, Yang
contents Multimodal large language models (MLLMs) are effective at capturing the semantics of short video content; however, they often fail to attend to the policy-specific details required for reliable content moderation. To address this limitation, we introduce IPS, a novel framework that integrates In-prompt Process Supervision into MLLMs by introducing sequential reasoning over ancillary questions during fine-tuning. IPS consistently outperforms baseline MLLMs on public and proprietary benchmarks. Moreover, replacing human-annotated ancillary labels with MLLM-generated ones results in only marginal performance degradation, demonstrating robustness to noisy supervision and strong scalability with model-generated annotations. These findings establish IPS as a scalable and effective solution for complex multimodal classification in large-scale industrial settings.
format Preprint
id arxiv_https___arxiv_org_abs_2412_15251
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle IPS: In-Prompt Process Supervision for Short Video Content Moderation
Liu, Mingchao
Sun, Yu
Sun, Ruixiao
Dong, Xin
Shen, Xiang
Wang, Hongwei
Xiong, Hongyu
Song, Yang
Computation and Language
Artificial Intelligence
Multimodal large language models (MLLMs) are effective at capturing the semantics of short video content; however, they often fail to attend to the policy-specific details required for reliable content moderation. To address this limitation, we introduce IPS, a novel framework that integrates In-prompt Process Supervision into MLLMs by introducing sequential reasoning over ancillary questions during fine-tuning. IPS consistently outperforms baseline MLLMs on public and proprietary benchmarks. Moreover, replacing human-annotated ancillary labels with MLLM-generated ones results in only marginal performance degradation, demonstrating robustness to noisy supervision and strong scalability with model-generated annotations. These findings establish IPS as a scalable and effective solution for complex multimodal classification in large-scale industrial settings.
title IPS: In-Prompt Process Supervision for Short Video Content Moderation
topic Computation and Language
Artificial Intelligence
url https://arxiv.org/abs/2412.15251