Skip to yearly menu bar Skip to main content


Poster

Interpreting CLIP with Sparse Linear Concept Embeddings (SpLiCE)

Usha Bhalla · Alex Oesterling · Suraj Srinivas · Flavio Calmon · Himabindu Lakkaraju

East Exhibit Hall A-C #3011
[ ]
Fri 13 Dec 4:30 p.m. PST — 7:30 p.m. PST

Abstract:

CLIP embeddings have demonstrated remarkable performance across a wide range of multimodal applications. However, these high-dimensional, dense vector representations are not easily interpretable, limiting our understanding of the rich structure of CLIP and its use in downstream applications that require transparency. In this work, we show that the semantic structure of CLIP's latent space can be leveraged to provide interpretability, allowing for the decomposition of representations into semantic concepts. We formulate this problem as one of sparse recovery and propose a novel method, Sparse Linear Concept Embeddings (SpLiCE), for transforming CLIP representations into sparse linear combinations of human-interpretable concepts. Distinct from previous work, \method is task-agnostic and can be used, without training, to explain and even replace traditional dense CLIP representations, maintaining high downstream performance while significantly improving their interpretability. We also demonstrate significant use cases of \method representations including detecting spurious correlations and model editing. Code is provided at https://github.com/AI4LIFE-GROUP/SpLiCE.

Live content is unavailable. Log in and register to view live content