Picture for Andreas Kirsch

Andreas Kirsch

All models are wrong, some are useful: Model Selection with Limited Labels

Add code
Oct 17, 2024
Viaarxiv icon

(Implicit) Ensembles of Ensembles: Epistemic Uncertainty Collapse in Large Models

Add code
Sep 04, 2024
Viaarxiv icon

Min P Sampling: Balancing Creativity and Coherence at High Temperature

Add code
Jul 01, 2024
Viaarxiv icon

The Benefits and Risks of Transductive Approaches for AI Fairness

Add code
Jun 17, 2024
Figure 1 for The Benefits and Risks of Transductive Approaches for AI Fairness
Figure 2 for The Benefits and Risks of Transductive Approaches for AI Fairness
Figure 3 for The Benefits and Risks of Transductive Approaches for AI Fairness
Figure 4 for The Benefits and Risks of Transductive Approaches for AI Fairness
Viaarxiv icon

CoLoR-Filter: Conditional Loss Reduction Filtering for Targeted Language Model Pre-training

Add code
Jun 15, 2024
Viaarxiv icon

Advancing Deep Active Learning & Data Subset Selection: Unifying Principles with Information-Theory Intuitions

Add code
Jan 09, 2024
Viaarxiv icon

Prediction-Oriented Bayesian Active Learning

Add code
Apr 17, 2023
Figure 1 for Prediction-Oriented Bayesian Active Learning
Figure 2 for Prediction-Oriented Bayesian Active Learning
Figure 3 for Prediction-Oriented Bayesian Active Learning
Figure 4 for Prediction-Oriented Bayesian Active Learning
Viaarxiv icon

Does `Deep Learning on a Data Diet' reproduce? Overall yes, but GraNd at Initialization does not

Add code
Mar 26, 2023
Viaarxiv icon

Black-Box Batch Active Learning for Regression

Add code
Feb 17, 2023
Viaarxiv icon

Speeding Up BatchBALD: A k-BALD Family of Approximations for Active Learning

Add code
Jan 23, 2023
Viaarxiv icon