Saved in:
Bibliographic Details
Main Authors: Jana, Soumyadeep, Danayak, Sahil, Singh, Sanasam Ranbir
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2507.04508
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917048739692544
author Jana, Soumyadeep
Danayak, Sahil
Singh, Sanasam Ranbir
author_facet Jana, Soumyadeep
Danayak, Sahil
Singh, Sanasam Ranbir
contents The growing prevalence of multimodal image-text sarcasm on social media poses challenges for opinion mining systems. Existing approaches rely on full fine-tuning of large models, making them unsuitable to adapt under resource-constrained settings. While recent parameter-efficient fine-tuning (PEFT) methods offer promise, their off-the-shelf use underperforms on complex tasks like sarcasm detection. We propose AdS-CLIP (Adapter-state Sharing in CLIP), a lightweight framework built on CLIP that inserts adapters only in the upper layers to preserve low-level unimodal representations in the lower layers and introduces a novel adapter-state sharing mechanism, where textual adapters guide visual ones to promote efficient cross-modal learning in the upper layers. Experiments on two public benchmarks demonstrate that AdS-CLIP not only outperforms standard PEFT methods but also existing multimodal baselines with significantly fewer trainable parameters.
format Preprint
id arxiv_https___arxiv_org_abs_2507_04508
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Adapter-state Sharing CLIP for Parameter-efficient Multimodal Sarcasm Detection
Jana, Soumyadeep
Danayak, Sahil
Singh, Sanasam Ranbir
Computation and Language
The growing prevalence of multimodal image-text sarcasm on social media poses challenges for opinion mining systems. Existing approaches rely on full fine-tuning of large models, making them unsuitable to adapt under resource-constrained settings. While recent parameter-efficient fine-tuning (PEFT) methods offer promise, their off-the-shelf use underperforms on complex tasks like sarcasm detection. We propose AdS-CLIP (Adapter-state Sharing in CLIP), a lightweight framework built on CLIP that inserts adapters only in the upper layers to preserve low-level unimodal representations in the lower layers and introduces a novel adapter-state sharing mechanism, where textual adapters guide visual ones to promote efficient cross-modal learning in the upper layers. Experiments on two public benchmarks demonstrate that AdS-CLIP not only outperforms standard PEFT methods but also existing multimodal baselines with significantly fewer trainable parameters.
title Adapter-state Sharing CLIP for Parameter-efficient Multimodal Sarcasm Detection
topic Computation and Language
url https://arxiv.org/abs/2507.04508