Support-Set Context Matters for Bongard Problems
This work improves machine learning performance on Bongard problems, a challenging benchmark for abstract reasoning, by incorporating support-set context, though it is incremental as it builds on existing vision backbones.
The paper tackled the problem of low accuracy in solving Bongard problems, which are IQ tests requiring abstract concept derivation from support images, by addressing the issue of methods not adapting image features based on the entire support set context, resulting in new state-of-the-art accuracies of 75.3% on Bongard-LOGO and 76.4% on Bongard-HOI.
Current machine learning methods struggle to solve Bongard problems, which are a type of IQ test that requires deriving an abstract "concept" from a set of positive and negative "support" images, and then classifying whether or not a new query image depicts the key concept. On Bongard-HOI, a benchmark for natural-image Bongard problems, most existing methods have reached at best 69% accuracy (where chance is 50%). Low accuracy is often attributed to neural nets' lack of ability to find human-like symbolic rules. In this work, we point out that many existing methods are forfeiting accuracy due to a much simpler problem: they do not adapt image features given information contained in the support set as a whole, and rely instead on information extracted from individual supports. This is a critical issue, because the "key concept" in a typical Bongard problem can often only be distinguished using multiple positives and multiple negatives. We explore simple methods to incorporate this context and show substantial gains over prior works, leading to new state-of-the-art accuracy on Bongard-LOGO (75.3%) and Bongard-HOI (76.4%) compared to methods with equivalent vision backbone architectures and strong performance on the original Bongard problem set (60.8%).