Long Context Question Answering via Supervised Contrastive Learning
This work addresses the challenge of evidence identification in long-context QA, which is crucial for accurate reasoning over lengthy documents, though it appears incremental as it builds on existing transformer models.
The authors tackled the problem of identifying supporting evidence in long-context question answering by proposing a supervised contrastive learning method to improve evidence discrimination, resulting in consistent improvements across three transformer models on HotpotQA and QAsper benchmarks.
Long-context question answering (QA) tasks require reasoning over a long document or multiple documents. Addressing these tasks often benefits from identifying a set of evidence spans (e.g., sentences), which provide supporting evidence for answering the question. In this work, we propose a novel method for equipping long-context QA models with an additional sequence-level objective for better identification of the supporting evidence. We achieve this via an additional contrastive supervision signal in finetuning, where the model is encouraged to explicitly discriminate supporting evidence sentences from negative ones by maximizing question-evidence similarity. The proposed additional loss exhibits consistent improvements on three different strong long-context transformer models, across two challenging question answering benchmarks -- HotpotQA and QAsper.