Enhancing Cognition and Explainability of Multimodal Foundation Models with Self-Synthesized Data

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Shi, Yucheng, Li, Quanzheng, Sun, Jin, Li, Xiang, Liu, Ninghao
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866910843045675008
author Shi, Yucheng
Li, Quanzheng
Sun, Jin
Li, Xiang
Liu, Ninghao
author_facet Shi, Yucheng
Li, Quanzheng
Sun, Jin
Li, Xiang
Liu, Ninghao
contents Large Multimodal Models (LMMs), or Vision-Language Models (VLMs), have shown impressive capabilities in a wide range of visual tasks. However, they often struggle with fine-grained visual reasoning, failing to identify domain-specific objectives and provide justifiable explanations for their predictions. To address the above challenge, we propose a novel visual rejection sampling framework to improve the cognition and explainability of LMMs using self-synthesized data. Specifically, visual fine-tuning requires images, queries, and target answers. Our approach begins by synthesizing interpretable answers that include human-verifiable visual features. These features are based on expert-defined concepts, and carefully selected based on their alignment with the image content. After each round of fine-tuning, we apply a reward model-free filtering mechanism to select the highest-quality interpretable answers for the next round of tuning. This iterative process of synthetic data generation and fine-tuning progressively improves the model's ability to generate accurate and reasonable explanations. Experimental results demonstrate the effectiveness of our method in improving both the accuracy and explainability of specialized visual classification tasks.
format Preprint
id arxiv_https___arxiv_org_abs_2502_14044
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Enhancing Cognition and Explainability of Multimodal Foundation Models with Self-Synthesized Data
Shi, Yucheng
Li, Quanzheng
Sun, Jin
Li, Xiang
Liu, Ninghao
Computer Vision and Pattern Recognition
Machine Learning
Large Multimodal Models (LMMs), or Vision-Language Models (VLMs), have shown impressive capabilities in a wide range of visual tasks. However, they often struggle with fine-grained visual reasoning, failing to identify domain-specific objectives and provide justifiable explanations for their predictions. To address the above challenge, we propose a novel visual rejection sampling framework to improve the cognition and explainability of LMMs using self-synthesized data. Specifically, visual fine-tuning requires images, queries, and target answers. Our approach begins by synthesizing interpretable answers that include human-verifiable visual features. These features are based on expert-defined concepts, and carefully selected based on their alignment with the image content. After each round of fine-tuning, we apply a reward model-free filtering mechanism to select the highest-quality interpretable answers for the next round of tuning. This iterative process of synthetic data generation and fine-tuning progressively improves the model's ability to generate accurate and reasonable explanations. Experimental results demonstrate the effectiveness of our method in improving both the accuracy and explainability of specialized visual classification tasks.
title Enhancing Cognition and Explainability of Multimodal Foundation Models with Self-Synthesized Data
topic Computer Vision and Pattern Recognition
Machine Learning
url https://arxiv.org/abs/2502.14044