Sharut Gupta
@sharut_gupta • 2,680 subscribers
PhD @MIT_CSAIL | Previously @GoogleDeepMind (Gemini), @AIatMeta | BTech @iitdelhi
Videos
![[1/n] Do distinct large models admit a simple map that aligns their embedding spaces? We show that across multimodal contrastive models—trained on different data and architectures—an orthogonal map aligns image embeddings. Strikingly, the same map also aligns text embeddings.](https://image.24vids.com/tw-2025925394046566905/amplify_video_thumb/2025924785780273152/img/JFykeWHF3Gm9flZw.jpg)
[1/n] Do distinct large models admit a simple map that aligns their embedding spaces? We show that across multimodal contrastive models—trained on different data and architectures—an orthogonal map aligns image embeddings. Strikingly, the same map also aligns text embeddings.
Sharut Gupta36,746 Aufrufe • vor 3 Monaten
Keine weiteren Inhalte verfügbar