---
title: "Unimodal Augmentation-based SSL Objectives Induce Cross-Modal Alignment"
source: "arXiv AI/CL/LG（论文）"
source_tier: "t2"
url: "https://arxiv.org/abs/2611.02917v1"
permalink: "https://feed.astralor.com/items/itm_f783435b9672a02767da0320"
published: "2026-11-02T10:49:23.000Z"
score: 56
channel: "ai"
tags: ["安全","多模态","Unimodal Augmentation-based SSL","Objectives Induce Cross-Modal","Alignment Recent","However","Recent","InfoNCE"]
---

# Unimodal Augmentation-based SSL Objectives Induce Cross-Modal Alignment

## At a glance

Recent works have observed implicit cross-modal alignment between independently trained unimodal models. However, the mechanisms underlying the emergence of this phenomenon remain unresolved. In this paper, we demonstra…

## Key points

- Recent works have observed implicit cross-modal alignment between independently trained unimodal models.
- However, the mechanisms underlying the emergence of this phenomenon remain unresolved.
- In this paper, we demonstrate that unimodal augmentation-based self-supervised objectives implicitly induce cross-modal alignment.

## Full text

Recent works have observed implicit cross-modal alignment between independently trained unimodal models. However, the mechanisms underlying the emergence of this phenomenon remain unresolved. In this paper, we demonstrate that unimodal augmentation-based self-supervised objectives implicitly induce cross-modal alignment. We show that low unimodal losses force high cross-modal alignment, whereas a high unimodal loss prevents such alignment. We theoretically connect the widely used InfoNCE loss to local CKA, the popular measure of cross-modal alignment. We further provide extensive empirical evidence supporting this claim, both in controlled settings and across 53 unconstrained frozen text, vision, and audio unimodal encoders. In both cases, unimodal augmentation-based objectives control cross-modal alignment. Overall, our results provide a theoretical and empirical foundation for understanding the emergence of implicit cross-modal alignment by establishing a direct connection between unimodal training objectives and cross-modal structure.

---

Verify against the original before citing; curated by Astral Feed: https://feed.astralor.com/items/itm_f783435b9672a02767da0320