Black-box access is insufficient for rigorous ai audits
External audits of AI systems are increasingly recognized as a key mechanism for AI
governance. The effectiveness of an audit, however, depends on the degree of access …
governance. The effectiveness of an audit, however, depends on the degree of access …
Sparse feature circuits: Discovering and editing interpretable causal graphs in language models
We introduce methods for discovering and applying sparse feature circuits. These are
causally implicated subnetworks of human-interpretable features for explaining language …
causally implicated subnetworks of human-interpretable features for explaining language …
Carzero: Cross-attention alignment for radiology zero-shot classification
Abstract The advancement of Zero-Shot Learning in the medical domain has been driven
forward by using pre-trained models on large-scale image-text pairs focusing on image-text …
forward by using pre-trained models on large-scale image-text pairs focusing on image-text …
Interpreting clip with sparse linear concept embeddings (splice)
CLIP embeddings have demonstrated remarkable performance across a wide range of
computer vision tasks. However, these high-dimensional, dense vector representations are …
computer vision tasks. However, these high-dimensional, dense vector representations are …
Decomposing and editing predictions by modeling model computation
How does the internal computation of a machine learning model transform inputs into
predictions? In this paper, we introduce a task called component modeling that aims to …
predictions? In this paper, we introduce a task called component modeling that aims to …
Erm++: An improved baseline for domain generalization
Domain Generalization (DG) measures a classifier's ability to generalize to new distributions
of data it was not trained on. Recent work has shown that a hyperparameter-tuned Empirical …
of data it was not trained on. Recent work has shown that a hyperparameter-tuned Empirical …
Diffusion Lens: Interpreting Text Encoders in Text-to-Image Pipelines
Text-to-image diffusion models (T2I) use a latent representation of a text prompt to guide the
image generation process. However, the process by which the encoder produces the text …
image generation process. However, the process by which the encoder produces the text …
LeGrad: An Explainability Method for Vision Transformers via Feature Formation Sensitivity
W Bousselham, A Boggust, S Chaybouti… - arXiv preprint arXiv …, 2024 - arxiv.org
Vision Transformers (ViTs), with their ability to model long-range dependencies through self-
attention mechanisms, have become a standard architecture in computer vision. However …
attention mechanisms, have become a standard architecture in computer vision. However …
Evolving Interpretable Visual Classifiers with Large Language Models
Multimodal pre-trained models, such as CLIP, are popular for zero-shot classification due to
their open-vocabulary flexibility and high performance. However, vision-language models …
their open-vocabulary flexibility and high performance. However, vision-language models …
Finding Visual Task Vectors
Visual Prompting is a technique for teaching models to perform a visual task via in-context
examples, without any additional training. In this work, we analyze the activations of MAE …
examples, without any additional training. In this work, we analyze the activations of MAE …