Picture for Thomas Fel

Thomas Fel

Local vs distributed representations: What is the right basis for interpretability?

Add code
Nov 06, 2024
Viaarxiv icon

Unearthing Skill-Level Insights for Understanding Trade-Offs of Foundation Models

Add code
Oct 17, 2024
Viaarxiv icon

One Wave to Explain Them All: A Unifying Perspective on Post-hoc Explainability

Add code
Oct 02, 2024
Viaarxiv icon

Understanding Visual Feature Reliance through the Lens of Complexity

Add code
Jul 08, 2024
Figure 1 for Understanding Visual Feature Reliance through the Lens of Complexity
Figure 2 for Understanding Visual Feature Reliance through the Lens of Complexity
Figure 3 for Understanding Visual Feature Reliance through the Lens of Complexity
Figure 4 for Understanding Visual Feature Reliance through the Lens of Complexity
Viaarxiv icon

Latent Representation Matters: Human-like Sketches in One-shot Drawing Tasks

Add code
Jun 10, 2024
Viaarxiv icon

Feature Accentuation: Revealing 'What' Features Respond to in Natural Images

Add code
Feb 15, 2024
Viaarxiv icon

On the Foundations of Shortcut Learning

Add code
Oct 24, 2023
Viaarxiv icon

Gradient strikes back: How filtering out high frequencies improves explanations

Add code
Jul 18, 2023
Figure 1 for Gradient strikes back: How filtering out high frequencies improves explanations
Figure 2 for Gradient strikes back: How filtering out high frequencies improves explanations
Figure 3 for Gradient strikes back: How filtering out high frequencies improves explanations
Figure 4 for Gradient strikes back: How filtering out high frequencies improves explanations
Viaarxiv icon

Unlocking Feature Visualization for Deeper Networks with MAgnitude Constrained Optimization

Add code
Jun 11, 2023
Figure 1 for Unlocking Feature Visualization for Deeper Networks with MAgnitude Constrained Optimization
Figure 2 for Unlocking Feature Visualization for Deeper Networks with MAgnitude Constrained Optimization
Figure 3 for Unlocking Feature Visualization for Deeper Networks with MAgnitude Constrained Optimization
Figure 4 for Unlocking Feature Visualization for Deeper Networks with MAgnitude Constrained Optimization
Viaarxiv icon

A Holistic Approach to Unifying Automatic Concept Extraction and Concept Importance Estimation

Add code
Jun 11, 2023
Figure 1 for A Holistic Approach to Unifying Automatic Concept Extraction and Concept Importance Estimation
Figure 2 for A Holistic Approach to Unifying Automatic Concept Extraction and Concept Importance Estimation
Figure 3 for A Holistic Approach to Unifying Automatic Concept Extraction and Concept Importance Estimation
Figure 4 for A Holistic Approach to Unifying Automatic Concept Extraction and Concept Importance Estimation
Viaarxiv icon