mlsanityData to 7 Oct 2026

Epsilon Sampling Rocks: Investigating Sampling Strategies for Minimum Bayes Risk Decoding for Machine Translation

arXiv 2305.09860 ↗First posted 17 May 2023cs.CL cs.AI cs.LGcited by 57

Not on the map. Why?

Abstract (arXiv)

Recent advances in machine translation (MT) have shown that Minimum Bayes Risk (MBR) decoding can be a powerful alternative to beam search decoding, especially when combined with neural-based utility functions. However, the performance of MBR decoding depends heavily on how and how many candidates are sampled from the model. In this paper, we explore how different sampling approaches for generating candidate lists for MBR decoding affect performance. We evaluate popular sampling approaches, such as ancestral, nucleus, and top-k sampling. Based on our insights into their limitations, we experiment with the recently proposed epsilon-sampling approach, which prunes away all tokens with a probability smaller than epsilon, ensuring that each token in a sample receives a fair probability mass. Through extensive human evaluations, we demonstrate that MBR decoding based on epsilon-sampling significantly outperforms not only beam search decoding, but also MBR decoding with all other tested sampling methods across four language pairs.

Key facts. Cited by 57 papers in the map (data through 7 Oct 2026). Builds on 19 of its 40 references that are in the map. First posted on arXiv 17 May 2023.

Builds on

19 of 40 references.

References · 19

  1. 2017Attention Is All You Needcited by 59,932
  2. 2014Neural Machine Translation by Jointly Learning to Align and Translatecited by 4,530
  3. 2019The Curious Case of Neural Text Degenerationcited by 2,881
  4. 2018SentencePiece: A simple and language independent subword tokenizer and detokenizer for Neural Text Processingcited by 1,973
  5. 2016Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translationcited by 1,471
  6. 2020COMET: A Neural Framework for MT Evaluationcited by 1,024
  7. 2017Convolutional Sequence to Sequence Learningcited by 550
  8. 2021Experts, Errors, and Context: A Large-Scale Study of Human Evaluation for Machine Translationcited by 333
  9. 2020Is MAP Decoding All You Need? The Inadequacy of the Mode in Neural Machine Translationcited by 137
  10. 2022Truncation Sampling as Language Model Desmoothingcited by 121
  11. 2020BLEU might be Guilty but References are not Innocentcited by 109
  12. 2021High Quality Rather than High Model Probability: Minimum Bayes Risk Decoding with Neural Metricscited by 104
7 more
  1. 2021A Statistical Analysis of Summarization Evaluation Metrics using Resampling Methodscited by 71
  2. 2021Understanding the Properties of Minimum Bayes Risk Decoding in Neural Machine Translationcited by 56
  3. 2022Quality-Aware Decoding for Neural Machine Translationcited by 49
  4. 2018Denoising Neural Machine Translation Training with Trusted Data and Online Data Selectioncited by 43
  5. 2022Identifying Weaknesses in Machine Translation Metrics Through Minimum Bayes Risk Decoding: A Case Study for COMETcited by 41
  6. 2020Learning to Evaluate Translation Beyond English: BLEURT Submissions to the WMT Metrics 2020 Shared Taskcited by 25
  7. 2022Adam Mickiewicz University at WMT 2022: NER-Assisted and Quality-Aware Neural Machine Translationcited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 9; 2024: 26; 2025: 14; 2026: 8.

57 papers.

Sort:

By area:

  1. 2024From Decoding to Meta-Generation: Inference-time Algorithms for Large Language Modelscited by 136
  2. 2024Ensemble Learning for Heterogeneous Large Language Models with Deep Parallel Collaborationcited by 66
  3. 2023Uncertainty in Natural Language Generation: From Theory to Applicationscited by 61
  4. 2023It's MBR All the Way Down: Modern Generation Techniques Through the Lens of Minimum Bayes Riskcited by 49
  5. 2023Faster Minimum Bayes Risk Decoding with Confidence-based Pruningcited by 34
  6. 2023MBR and QE Finetuning: Training-time Distillation of the Best and Most Expensive Decoding Methodscited by 33
  7. 2023Closing the Curious Case of Neural Text Degenerationcited by 30
  8. 2023Direct Preference Optimization for Neural Machine Translation with Minimum Bayes Risk Decodingcited by 28
  9. 2024Non-Exchangeable Conformal Language Generation with Nearest Neighborscited by 21
  10. 2024Linear-time Minimum Bayes Risk Decoding with Reference Aggregationcited by 21
  11. 2024Better Instruction-Following Through Minimum Bayes Riskcited by 19
  12. 2024Narrowing the Knowledge Evaluation Gap: Open-Domain Question Answering with Multi-Granularity Answerscited by 18
38 more
  1. 2024QUEST: Quality-Aware Metropolis-Hastings Sampling for Machine Translationcited by 16
  2. 2024Hyperparameter-Free Approach for Faster Minimum Bayes Risk Decodingcited by 15
  3. 2024Mitigating Metric Bias in Minimum Bayes Risk Decodingcited by 15
  4. 2024Don't Rank, Combine! Combining Machine Translation Hypotheses Using Quality Estimationcited by 13
  5. 2024Regression-aware Inference with LLMscited by 13
  6. 2024Efficient Minimum Bayes Risk Decoding using Low-Rank Matrix Completion Algorithmscited by 13
  7. 2024Improving Minimum Bayes Risk Decoding with Multi-Promptcited by 13
  8. 2024Centroid-Based Efficient Minimum Bayes Risk Decodingcited by 12
  9. 2024Generating Diverse and High-Quality Texts by Minimum Bayes Risk Decodingcited by 11
  10. 2023Model-Based Minimum Bayes Risk Decoding for Text Generationcited by 10
  11. 2024Introducing the NewsPaLM MBR and QE Dataset: LLM-Generated High-Quality Parallel Data Outperforms Traditional Web-Crawled Datacited by 10
  12. 2023Quality-Aware Translation Models: Efficient Generation and Quality Estimation in a Single Modelcited by 9
  13. 2025When Life Gives You Samples: The Benefits of Scaling up Inference Compute for Multilingual LLMscited by 9
  14. 2025Semantic-guided Diverse Decoding for Large Language Modelcited by 9
  15. 2024Diversity Explains Inference Scaling Laws: Through a Case Study of Minimum Bayes Risk Decodingcited by 8
  16. 2024Chasing COMET: Leveraging Minimum Bayes Risk Decoding for Self-Improving Machine Translationcited by 6
  17. 2024SpecFuse: Ensembling Large Language Models via Next-Segment Predictioncited by 6
  18. 2025Mitigating Hallucinated Translations in Large Language Models with Hallucination-focused Preference Optimizationcited by 6
  19. 2025Adding Chocolate to Mint: Mitigating Metric Interference in Machine Translationcited by 6
  20. 2024EMTeC: A Corpus of Eye Movements on Machine-Generated Textscited by 5
  21. 2025Representation Consistency for Accurate and Coherent LLM Answer Aggregationcited by 5
  22. 2024Enhancing Perception Capabilities of Multimodal LLMs with Training-Free Fusioncited by 4
  23. 2025Translate Smart, not Hard: Cascaded Translation Systems with Quality-Aware Deferralcited by 4
  24. 2025p-less Sampling: A Robust Hyperparameter-Free Approach for LLM Decodingcited by 4
  25. 2024Annotation-Efficient Language Model Alignment via Diverse and Representative Response Textscited by 3
  26. 2025Theoretical Guarantees for Minimum Bayes Risk Decodingcited by 3
  27. 2024On the True Distribution Approximation of Minimum Bayes-Risk Decodingcited by 2
  28. 2024A Bayesian Optimization Approach to Machine Translation Rerankingcited by 2
  29. 2024A Context-aware Framework for Translation-mediated Conversationscited by 2
  30. 2025Re-evaluating Minimum Bayes Risk Decoding for Automatic Speech Recognitioncited by 2
  31. 2025Agreement-Constrained Probabilistic Minimum Bayes Risk Decodingcited by 2
  32. 2026Task-Awareness Improves LLM Generations and Uncertaintycited by 2
  33. 2026One Single Hub Text Breaks CLIP: Identifying Vulnerabilities in Cross-Modal Encoders via Hubnesscited by 2
  34. 2023Separating the Wheat from the Chaff with BREAD: An open-source benchmark and metrics to detect redundancy in textcited by 1
  35. 2025Case-Based Decision-Theoretic Decoding with Quality Memoriescited by 1
  36. 2025Quality Estimation Reranking for Document-Level Translationcited by 1
  37. 2026Decoding in Geometry: Alleviating Embedding-Space Crowding for Complex Reasoningcited by 1
  38. 2026Breaking the Likelihood Trap: Variance-Calibrated Modulation for Large Language Model Decodingcited by 1

50 most cited of 57 All 57 (JSON)