Picture for Piotr Mardziel

Piotr Mardziel

De-amplifying Bias from Differential Privacy in Language Model Fine-tuning

Add code
Feb 07, 2024
Viaarxiv icon

Abstracting Influence Paths for Explaining (Contextualization of) BERT Models

Add code
Nov 02, 2020
Figure 1 for Abstracting Influence Paths for Explaining (Contextualization of) BERT Models
Figure 2 for Abstracting Influence Paths for Explaining (Contextualization of) BERT Models
Figure 3 for Abstracting Influence Paths for Explaining (Contextualization of) BERT Models
Figure 4 for Abstracting Influence Paths for Explaining (Contextualization of) BERT Models
Viaarxiv icon

Towards Behavior-Level Explanation for Deep Reinforcement Learning

Add code
Sep 17, 2020
Figure 1 for Towards Behavior-Level Explanation for Deep Reinforcement Learning
Figure 2 for Towards Behavior-Level Explanation for Deep Reinforcement Learning
Figure 3 for Towards Behavior-Level Explanation for Deep Reinforcement Learning
Figure 4 for Towards Behavior-Level Explanation for Deep Reinforcement Learning
Viaarxiv icon

Fairness Under Feature Exemptions: Counterfactual and Observational Measures

Add code
Jun 14, 2020
Figure 1 for Fairness Under Feature Exemptions: Counterfactual and Observational Measures
Figure 2 for Fairness Under Feature Exemptions: Counterfactual and Observational Measures
Figure 3 for Fairness Under Feature Exemptions: Counterfactual and Observational Measures
Figure 4 for Fairness Under Feature Exemptions: Counterfactual and Observational Measures
Viaarxiv icon

Smoothed Geometry for Robust Attribution

Add code
Jun 11, 2020
Figure 1 for Smoothed Geometry for Robust Attribution
Figure 2 for Smoothed Geometry for Robust Attribution
Figure 3 for Smoothed Geometry for Robust Attribution
Figure 4 for Smoothed Geometry for Robust Attribution
Viaarxiv icon

Influence Paths for Characterizing Subject-Verb Number Agreement in LSTM Language Models

Add code
May 03, 2020
Figure 1 for Influence Paths for Characterizing Subject-Verb Number Agreement in LSTM Language Models
Figure 2 for Influence Paths for Characterizing Subject-Verb Number Agreement in LSTM Language Models
Figure 3 for Influence Paths for Characterizing Subject-Verb Number Agreement in LSTM Language Models
Figure 4 for Influence Paths for Characterizing Subject-Verb Number Agreement in LSTM Language Models
Viaarxiv icon

Gender Bias in Neural Natural Language Processing

Add code
Jul 31, 2018
Figure 1 for Gender Bias in Neural Natural Language Processing
Figure 2 for Gender Bias in Neural Natural Language Processing
Figure 3 for Gender Bias in Neural Natural Language Processing
Figure 4 for Gender Bias in Neural Natural Language Processing
Viaarxiv icon

Supervising Feature Influence

Add code
Apr 07, 2018
Figure 1 for Supervising Feature Influence
Figure 2 for Supervising Feature Influence
Figure 3 for Supervising Feature Influence
Viaarxiv icon

Use Privacy in Data-Driven Systems: Theory and Experiments with Machine Learnt Programs

Add code
Sep 07, 2017
Figure 1 for Use Privacy in Data-Driven Systems: Theory and Experiments with Machine Learnt Programs
Figure 2 for Use Privacy in Data-Driven Systems: Theory and Experiments with Machine Learnt Programs
Figure 3 for Use Privacy in Data-Driven Systems: Theory and Experiments with Machine Learnt Programs
Figure 4 for Use Privacy in Data-Driven Systems: Theory and Experiments with Machine Learnt Programs
Viaarxiv icon

Proxy Non-Discrimination in Data-Driven Systems

Add code
Jul 25, 2017
Figure 1 for Proxy Non-Discrimination in Data-Driven Systems
Figure 2 for Proxy Non-Discrimination in Data-Driven Systems
Figure 3 for Proxy Non-Discrimination in Data-Driven Systems
Figure 4 for Proxy Non-Discrimination in Data-Driven Systems
Viaarxiv icon