On the Transferability of Minimal Prediction Preserving Inputs in Question Answering
This work addresses the problem of understanding model generalization and interpretability for researchers in NLP and machine learning, revealing limitations in using MPPIs as a diagnostic tool.
The paper investigates Minimal Prediction Preserving Inputs (MPPIs) in question answering models, finding that these short, uninterpretable fragments are highly transferable across domains and invariant to training variations, achieving significantly higher performance than comparable queries.
Recent work (Feng et al., 2018) establishes the presence of short, uninterpretable input fragments that yield high confidence and accuracy in neural models. We refer to these as Minimal Prediction Preserving Inputs (MPPIs). In the context of question answering, we investigate competing hypotheses for the existence of MPPIs, including poor posterior calibration of neural models, lack of pretraining, and "dataset bias" (where a model learns to attend to spurious, non-generalizable cues in the training data). We discover a perplexing invariance of MPPIs to random training seed, model architecture, pretraining, and training domain. MPPIs demonstrate remarkable transferability across domains achieving significantly higher performance than comparably short queries. Additionally, penalizing over-confidence on MPPIs fails to improve either generalization or adversarial robustness. These results suggest the interpretability of MPPIs is insufficient to characterize generalization capacity of these models. We hope this focused investigation encourages more systematic analysis of model behavior outside of the human interpretable distribution of examples.