Frozen LVLMs for Micro-Video Recommendation: A Systematic Study of Feature Extraction and Fusion

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Sun, Huatuan, Ma, Yunshan, Wu, Changguang, Zhang, Yanxin, Wang, Pengfei, Du, Xiaoyu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917459950305280
author Sun, Huatuan
Ma, Yunshan
Wu, Changguang
Zhang, Yanxin
Wang, Pengfei
Du, Xiaoyu
author_facet Sun, Huatuan
Ma, Yunshan
Wu, Changguang
Zhang, Yanxin
Wang, Pengfei
Du, Xiaoyu
contents Frozen Large Video Language Models (LVLMs) are increasingly employed in micro-video recommendation due to their strong multimodal understanding. However, their integration lacks systematic empirical evaluation: practitioners typically deploy LVLMs as fixed black-box feature extractors without systematically comparing alternative representation strategies. To address this gap, we present the first systematic empirical study along two key design dimensions: (i) integration strategies with ID embeddings, specifically replacement versus fusion, and (ii) feature extraction paradigms, comparing LVLM-generated captions with intermediate decoder hidden states. Extensive experiments on representative LVLMs reveal three key principles: (1) intermediate hidden states consistently outperform caption-based representations, as natural-language summarization inevitably discards fine-grained visual semantics crucial for recommendation; (2) ID embeddings capture irreplaceable collaborative signals, rendering fusion strictly superior to replacement; and (3) the effectiveness of intermediate decoder features varies significantly across layers. Guided by these insights, we propose the Dual Feature Fusion (DFF) Framework, a lightweight and plug-and-play approach that adaptively fuses multi-layer representations from frozen LVLMs with item ID embeddings. DFF achieves state-of-the-art performance on two real-world micro-video recommendation benchmarks, consistently outperforming strong baselines and providing a principled approach to integrating off-the-shelf large vision-language models into micro-video recommender systems.
format Preprint
id arxiv_https___arxiv_org_abs_2512_21863
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Frozen LVLMs for Micro-Video Recommendation: A Systematic Study of Feature Extraction and Fusion
Sun, Huatuan
Ma, Yunshan
Wu, Changguang
Zhang, Yanxin
Wang, Pengfei
Du, Xiaoyu
Information Retrieval
Multimedia
Frozen Large Video Language Models (LVLMs) are increasingly employed in micro-video recommendation due to their strong multimodal understanding. However, their integration lacks systematic empirical evaluation: practitioners typically deploy LVLMs as fixed black-box feature extractors without systematically comparing alternative representation strategies. To address this gap, we present the first systematic empirical study along two key design dimensions: (i) integration strategies with ID embeddings, specifically replacement versus fusion, and (ii) feature extraction paradigms, comparing LVLM-generated captions with intermediate decoder hidden states. Extensive experiments on representative LVLMs reveal three key principles: (1) intermediate hidden states consistently outperform caption-based representations, as natural-language summarization inevitably discards fine-grained visual semantics crucial for recommendation; (2) ID embeddings capture irreplaceable collaborative signals, rendering fusion strictly superior to replacement; and (3) the effectiveness of intermediate decoder features varies significantly across layers. Guided by these insights, we propose the Dual Feature Fusion (DFF) Framework, a lightweight and plug-and-play approach that adaptively fuses multi-layer representations from frozen LVLMs with item ID embeddings. DFF achieves state-of-the-art performance on two real-world micro-video recommendation benchmarks, consistently outperforming strong baselines and providing a principled approach to integrating off-the-shelf large vision-language models into micro-video recommender systems.
title Frozen LVLMs for Micro-Video Recommendation: A Systematic Study of Feature Extraction and Fusion
topic Information Retrieval
Multimedia
url https://arxiv.org/abs/2512.21863