Do Question Answering Modeling Improvements Hold Across Benchmarks?
This addresses the problem of benchmark reliability for researchers in NLP, showing that modeling improvements hold broadly across benchmarks, though it is incremental in validating existing practices.
The study investigated whether improvements in question answering models generalize across diverse benchmarks by introducing concurrence to measure ranking similarity, finding that human-constructed, downsampled, and programmatically-generated benchmarks show high concurrence, indicating broad applicability of modeling improvements.
Do question answering (QA) modeling improvements (e.g., choice of architecture and training procedure) hold consistently across the diverse landscape of QA benchmarks? To study this question, we introduce the notion of concurrence -- two benchmarks have high concurrence on a set of modeling approaches if they rank the modeling approaches similarly. We measure the concurrence between 32 QA benchmarks on a set of 20 diverse modeling approaches and find that human-constructed benchmarks have high concurrence amongst themselves, even if their passage and question distributions are very different. Surprisingly, even downsampled human-constructed benchmarks (i.e., collecting less data) and programmatically-generated benchmarks (e.g., cloze-formatted examples) have high concurrence with human-constructed benchmarks. These results indicate that, despite years of intense community focus on a small number of benchmarks, the modeling improvements studied hold broadly.