Open-Set Recognition Interaction Effects: Modular Gains and Where to Find Them

ICLR 2026 Conference Submission21533 Authors

Keywords: open-set recognition, interaction effects, auxiliary data, feature magnitude, large-scale evaluation

Abstract:

Open-set recognition (OSR) requires neural networks to classify known classes while rejecting unknown samples, which is critical for real-world deployment. So far, OSR research studied and developed representation learning and postprocessing methods independently and their interaction effects remain unexplored, leaving potential performance gains untapped. In this paper, we present the first systematic study of these interactions across dataset scales and auxiliary data usage. First, we discover a failure mode we term magnitude collapse, where representation learning methods that utilize auxiliary data can suffer performance degradation at large scale and irreversibly destroy discriminative information, despite excelling at small scale. Second, we study the interaction effects between representation learning and postprocessing methods, and reveal when they can be leveraged for modular performance gains via two-stage processing.
We also show where interaction effects amplify performance degradation due to magnitude collapse. Third, we show how these findings can be used to achieve state-of-the-art performance with a simple baseline and two-stage processing of OSR techniques. Finally, our results demonstrate that small-scale evaluations with auxiliary data are not predictive of large-scale performance, invalidating current best practices in OSR research.

Primary Area: unsupervised, self-supervised, semi-supervised, and supervised representation learning

Submission Number: 21533