Vision-Language Models Create Cross-Modal Task Representations

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Luo, Grace, Darrell, Trevor, Bar, Amir
Format: Preprint
Published: 2024
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912364262064128
author Luo, Grace
Darrell, Trevor
Bar, Amir
author_facet Luo, Grace
Darrell, Trevor
Bar, Amir
contents Autoregressive vision-language models (VLMs) can handle many tasks within a single model, yet the representations that enable this capability remain opaque. We find that VLMs align conceptually equivalent inputs into a shared task vector, which is invariant to modality (text, image) and format (examples, instruction), and may simplify VLM processing. We measure this alignment via cross-modal transfer -- the ability of a task vector derived in one modality to trigger the correct generation in another -- on a range of tasks and model architectures. Although the task vector is highly compressed, we find that this single vector outperforms prompting the model with the full task information, unique to this cross-modal case. Furthermore, we show that task vectors can be transferred from a base language model to its fine-tuned vision-language counterpart, and that they can be derived solely from instructions without the need for examples. Taken together, our findings shed light on how VLMs internally process task information, and how they map different modalities into common semantic representations. Project page: https://vlm-cross-modal-reps.github.io.
format Preprint
id arxiv_https___arxiv_org_abs_2410_22330
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Vision-Language Models Create Cross-Modal Task Representations
Luo, Grace
Darrell, Trevor
Bar, Amir
Computer Vision and Pattern Recognition
Computation and Language
Machine Learning
Autoregressive vision-language models (VLMs) can handle many tasks within a single model, yet the representations that enable this capability remain opaque. We find that VLMs align conceptually equivalent inputs into a shared task vector, which is invariant to modality (text, image) and format (examples, instruction), and may simplify VLM processing. We measure this alignment via cross-modal transfer -- the ability of a task vector derived in one modality to trigger the correct generation in another -- on a range of tasks and model architectures. Although the task vector is highly compressed, we find that this single vector outperforms prompting the model with the full task information, unique to this cross-modal case. Furthermore, we show that task vectors can be transferred from a base language model to its fine-tuned vision-language counterpart, and that they can be derived solely from instructions without the need for examples. Taken together, our findings shed light on how VLMs internally process task information, and how they map different modalities into common semantic representations. Project page: https://vlm-cross-modal-reps.github.io.
title Vision-Language Models Create Cross-Modal Task Representations
topic Computer Vision and Pattern Recognition
Computation and Language
Machine Learning
url https://arxiv.org/abs/2410.22330