Noisy-Channel Minimum Bayes Risk Decoding
2026-07-06 • Machine Learning
Machine LearningArtificial IntelligenceComputation and Language
AI summaryⓘ
The authors study Minimum Bayes Risk (MBR) decoding, which helps create better text by picking sentences that work well compared to example sentences. They point out that common ways to measure quality, like BLEU and COMET, look at comparisons in one direction only, which can cause problems. To fix this, the authors break down MBR decoding into parts that consider both directions between sentences and references. Their approach helps explain different MBR methods and shows that adjusting how much each part counts could improve text generation.
Minimum Bayes Risk (MBR) decodingmaximum a posteriori (MAP) decodingBLEUCOMETexpected utilitynoisy channel modelhypothesisreferencetext generationasymmetric evaluation metrics
Authors
Yusuke Sakai, Hidetaka Kamigaito, Taro Watanabe
Abstract
Minimum Bayes Risk (MBR) decoding yields more robust and higher-quality text generation than maximum a posteriori (MAP) decoding by selecting hypotheses that maximize expected utility over sampled pseudo-references. However, there exists a discrepancy in the design: hypothesis selection calculates expected utility scores conditioned on given pseudo-references, while commonly used evaluation metrics, e.g., BLEU and COMET, are asymmetric. Therefore, it is important to consider both hypothesis-to-reference and reference-to-hypothesis directional effects. In this study, we introduce a noisy channel decomposition of MBR decoding that naturally incorporates bidirectional effects to account for these asymmetries. We decompose MBR decoding into four interacting components: hypothesis-to-reference likelihood, reference-to-hypothesis likelihood, hypothesis prior, and reference prior. This decomposition provides a unified interpretation of existing MBR variants and enables metric- and task-specific interpretability by isolating the contribution of each channel. Our comprehensive analysis reveals that channel-wise contributions exhibit distinct characteristics across metrics while remaining consistent across tasks, and suggests that appropriate channel weighting may lead to improvements over original MBR decoding.