Sparse autoencoders reveal selective remapping of visual concepts during adaptation

• URL: http://arxiv.org/abs/2412.05276v1
• Date: Fri, 06 Dec 2024 18:59:51 GMT
• Title: Sparse autoencoders reveal selective remapping of visual concepts during adaptation
• Authors: Hyesu Lim, Jinho Choi, Jaegul Choo, Steffen Schneider,
• Abstract summary: Adapting foundation models for specific purposes has become a standard approach to build machine learning systems. We develop a new Sparse Autoencoder (SAE) for the CLIP vision transformer, named PatchSAE, to extract interpretable concepts.
• Score: 54.82630842681845
• License:
• Abstract: Adapting foundation models for specific purposes has become a standard approach to build machine learning systems for downstream applications. Yet, it is an open question which mechanisms take place during adaptation. Here we develop a new Sparse Autoencoder (SAE) for the CLIP vision transformer, named PatchSAE, to extract interpretable concepts at granular levels (e.g. shape, color, or semantics of an object) and their patch-wise spatial attributions. We explore how these concepts influence the model output in downstream image classification tasks and investigate how recent state-of-the-art prompt-based adaptation techniques change the association of model inputs to these concepts. While activations of concepts slightly change between adapted and non-adapted models, we find that the majority of gains on common adaptation tasks can be explained with the existing concepts already present in the non-adapted foundation model. This work provides a concrete framework to train and use SAEs for Vision Transformers and provides insights into explaining adaptation mechanisms.

Related papers

• Concept Layers: Enhancing Interpretability and Intervenability via LLM Conceptualization [2.163881720692685]
  We introduce a new methodology for incorporating interpretability and intervenability into an existing model by integrating Concept Layers into its architecture. Our approach projects the model's internal vector representations into a conceptual, explainable vector space before reconstructing and feeding them back into the model. We evaluate CLs across multiple tasks, demonstrating that they maintain the original model's performance and agreement while enabling meaningful interventions.
  arXiv  Detail & Related papers  (2025-02-19T11:10:19Z)
• Analyzing Fine-tuning Representation Shift for Multimodal LLMs Steering alignment [53.90425382758605]
  We show how fine-tuning alters the internal structure of a model to specialize in new multimodal tasks. Our work sheds light on how multimodal representations evolve through fine-tuning and offers a new perspective for interpreting model adaptation in multimodal tasks.
  arXiv  Detail & Related papers  (2025-01-06T13:37:13Z)
• Emergence of Abstractions: Concept Encoding and Decoding Mechanism for In-Context Learning in Transformers [18.077009146950473]
  Autoregressive transformers exhibit adaptive learning through in-context learning (ICL) We propose concept encoding-decoding mechanism to explain ICL by studying how transformers form and use internal abstractions in their representations. Our empirical insights shed light into better understanding the success and failure modes of large language models via their representations.
  arXiv  Detail & Related papers  (2024-12-16T19:00:18Z)
• Enforcing Interpretability in Time Series Transformers: A Concept Bottleneck Framework [2.8470354623829577]
  We develop a framework based on Concept Bottleneck Models to enforce interpretability of time series Transformers. We modify the training objective to encourage a model to develop representations similar to predefined interpretable concepts. We find that the model performance remains mostly unaffected, while the model shows much improved interpretability.
  arXiv  Detail & Related papers  (2024-10-08T14:22:40Z)
• Improving the Explain-Any-Concept by Introducing Nonlinearity to the Trainable Surrogate Model [4.6040036610482655]
  Explain Any Concept (EAC) model is a flexible method for explaining decisions. EAC model is based on using a surrogate model which has one trainable linear layer to simulate the target model. We show that by introducing an additional nonlinear layer to the original surrogate model, we can improve the performance of the EAC model.
  arXiv  Detail & Related papers  (2024-05-20T07:25:09Z)
• Improving Intervention Efficacy via Concept Realignment in Concept Bottleneck Models [57.86303579812877]
  Concept Bottleneck Models (CBMs) ground image classification on human-understandable concepts to allow for interpretable model decisions. Existing approaches often require numerous human interventions per image to achieve strong performances. We introduce a trainable concept realignment intervention module, which leverages concept relations to realign concept assignments post-intervention.
  arXiv  Detail & Related papers  (2024-05-02T17:59:01Z)
• Condition-Invariant Semantic Segmentation [77.10045325743644]
  We implement Condition-Invariant Semantic (CISS) on the current state-of-the-art domain adaptation architecture. Our method achieves the second-best performance on the normal-to-adverse Cityscapes$to$ACDC benchmark. CISS is shown to generalize well to domains unseen during training, such as BDD100K-night and ACDC-night.
  arXiv  Detail & Related papers  (2023-05-27T03:05:07Z)
• Vision Transformer with Quadrangle Attention [76.35955924137986]
  We propose a novel quadrangle attention (QA) method that extends the window-based attention to a general quadrangle formulation. Our method employs an end-to-end learnable quadrangle regression module that predicts a transformation matrix to transform default windows into target quadrangles. We integrate QA into plain and hierarchical vision transformers to create a new architecture named QFormer, which offers minor code modifications and negligible extra computational cost.
  arXiv  Detail & Related papers  (2023-03-27T11:13:50Z)
• FLAT: Few-Shot Learning via Autoencoding Transformation Regularizers [67.46036826589467]
  We present a novel regularization mechanism by learning the change of feature representations induced by a distribution of transformations without using the labels of data examples. It could minimize the risk of overfitting into base categories by inspecting the transformation-augmented variations at the encoded feature level. Experiment results show the superior performances to the current state-of-the-art methods in literature.
  arXiv  Detail & Related papers  (2019-12-29T15:26:28Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.