[ICLR 2025] - Cross the Gap: Exposing the Intra-modal Misalignment in CLIP via Modality Inversion

A repository for visualization of modality gap in VLMs

This repository contains the implementation of a modified LLaVA architecture designed to address information imbalance between modalities in multimodal learning.

Code repository for "Post-pre-training for Modality Alignment in Vision-Language Foundation Models" (CVPR2025)