- FaithEval_ Can Your Language Model Stay Faithful to Context, Even If _The Moon is Made of Marshmallows_
- _What Data Benefits My Classifier__ Enhancing Model Performance and Interpretability through Influence-Based Data Selection
- _Who experiences large model decay and why__ A Hierarchical Framework for Diagnosing Heterogeneous Performance Drift
- _Why Is There a Tumor___ Tell Me the Reason, Show Me the Evidence
- _Why did the Model Fail___ Attributing Model Performance Changes to Distribution Shifts
-
4.04 kB
-
73 Bytes