Dreaming Out Loud: A Self-Synthesis Approach For Training Vision-Language Models With Developmentally Plausible Data

Fuente: arXiv
Enregistré dans:
Détails bibliographiques
Auteurs principaux: AlKhamissi, Badr, Tang, Yingtian, Gökce, Abdülkadir, Mehrer, Johannes, Schrimpf, Martin
Format: Preprint
Publié: 2024
Sujets:
Accès en ligne:
Tags: Ajouter un tag
Pas de tags, Soyez le premier à ajouter un tag!
_version_ 1866917826750578688
author AlKhamissi, Badr
Tang, Yingtian
Gökce, Abdülkadir
Mehrer, Johannes
Schrimpf, Martin
author_facet AlKhamissi, Badr
Tang, Yingtian
Gökce, Abdülkadir
Mehrer, Johannes
Schrimpf, Martin
contents While today's large language models exhibit impressive abilities in generating human-like text, they require massive amounts of data during training. We here take inspiration from human cognitive development to train models in limited data conditions. Specifically we present a self-synthesis approach that iterates through four phases: Phase 1 sets up fundamental language abilities, training the model from scratch on a small corpus. Language is then associated with the visual environment in phase 2, integrating the model with a vision encoder to generate descriptive captions from labeled images. In the "self-synthesis" phase 3, the model generates captions for unlabeled images, that it then uses to further train its language component with a mix of synthetic, and previous real-world text. This phase is meant to expand the model's linguistic repertoire, similar to humans self-annotating new experiences. Finally, phase 4 develops advanced cognitive skills, by training the model on specific tasks such as visual question answering and reasoning. Our approach offers a proof of concept for training a multimodal model using a developmentally plausible amount of data.
format Preprint
id arxiv_https___arxiv_org_abs_2411_00828
institution arXiv
publishDate 2024
record_format arxiv
spellingShingle Dreaming Out Loud: A Self-Synthesis Approach For Training Vision-Language Models With Developmentally Plausible Data
AlKhamissi, Badr
Tang, Yingtian
Gökce, Abdülkadir
Mehrer, Johannes
Schrimpf, Martin
Computer Vision and Pattern Recognition
Machine Learning
While today's large language models exhibit impressive abilities in generating human-like text, they require massive amounts of data during training. We here take inspiration from human cognitive development to train models in limited data conditions. Specifically we present a self-synthesis approach that iterates through four phases: Phase 1 sets up fundamental language abilities, training the model from scratch on a small corpus. Language is then associated with the visual environment in phase 2, integrating the model with a vision encoder to generate descriptive captions from labeled images. In the "self-synthesis" phase 3, the model generates captions for unlabeled images, that it then uses to further train its language component with a mix of synthetic, and previous real-world text. This phase is meant to expand the model's linguistic repertoire, similar to humans self-annotating new experiences. Finally, phase 4 develops advanced cognitive skills, by training the model on specific tasks such as visual question answering and reasoning. Our approach offers a proof of concept for training a multimodal model using a developmentally plausible amount of data.
title Dreaming Out Loud: A Self-Synthesis Approach For Training Vision-Language Models With Developmentally Plausible Data
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2411.00828