Model-Based Minimum Bayes Risk Decoding for Text Generation

Yuu Jinnai, Tetsuro Morimura, Ukyo Honda, Kaito Ariu, Kenshi Abe
Proceedings of the 41st International Conference on Machine Learning, PMLR 235:22326-22347, 2024.

Abstract

Minimum Bayes Risk (MBR) decoding has been shown to be a powerful alternative to beam search decoding in a variety of text generation tasks. MBR decoding selects a hypothesis from a pool of hypotheses that has the least expected risk under a probability model according to a given utility function. Since it is impractical to compute the expected risk exactly over all possible hypotheses, two approximations are commonly used in MBR. First, it integrates over a sampled set of hypotheses rather than over all possible hypotheses. Second, it estimates the probability of each hypothesis using a Monte Carlo estimator. While the first approximation is necessary to make it computationally feasible, the second is not essential since we typically have access to the model probability at inference time. We propose model-based MBR (MBMBR), a variant of MBR that uses the model probability itself as the estimate of the probability distribution instead of the Monte Carlo estimate. We show analytically and empirically that the model-based estimate is more promising than the Monte Carlo estimate in text generation tasks. Our experiments show that MBMBR outperforms MBR in several text generation tasks, both with encoder-decoder models and with language models.

Cite this Paper

BibTeX
@InProceedings{pmlr-v235-jinnai24a, title = {Model-Based Minimum {B}ayes Risk Decoding for Text Generation}, author = {Jinnai, Yuu and Morimura, Tetsuro and Honda, Ukyo and Ariu, Kaito and Abe, Kenshi}, booktitle = {Proceedings of the 41st International Conference on Machine Learning}, pages = {22326--22347}, year = {2024}, editor = {Salakhutdinov, Ruslan and Kolter, Zico and Heller, Katherine and Weller, Adrian and Oliver, Nuria and Scarlett, Jonathan and Berkenkamp, Felix}, volume = {235}, series = {Proceedings of Machine Learning Research}, month = {21--27 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v235/main/assets/jinnai24a/jinnai24a.pdf}, url = {https://proceedings.mlr.press/v235/jinnai24a.html}, abstract = {Minimum Bayes Risk (MBR) decoding has been shown to be a powerful alternative to beam search decoding in a variety of text generation tasks. MBR decoding selects a hypothesis from a pool of hypotheses that has the least expected risk under a probability model according to a given utility function. Since it is impractical to compute the expected risk exactly over all possible hypotheses, two approximations are commonly used in MBR. First, it integrates over a sampled set of hypotheses rather than over all possible hypotheses. Second, it estimates the probability of each hypothesis using a Monte Carlo estimator. While the first approximation is necessary to make it computationally feasible, the second is not essential since we typically have access to the model probability at inference time. We propose model-based MBR (MBMBR), a variant of MBR that uses the model probability itself as the estimate of the probability distribution instead of the Monte Carlo estimate. We show analytically and empirically that the model-based estimate is more promising than the Monte Carlo estimate in text generation tasks. Our experiments show that MBMBR outperforms MBR in several text generation tasks, both with encoder-decoder models and with language models.} }
Endnote
%0 Conference Paper %T Model-Based Minimum Bayes Risk Decoding for Text Generation %A Yuu Jinnai %A Tetsuro Morimura %A Ukyo Honda %A Kaito Ariu %A Kenshi Abe %B Proceedings of the 41st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2024 %E Ruslan Salakhutdinov %E Zico Kolter %E Katherine Heller %E Adrian Weller %E Nuria Oliver %E Jonathan Scarlett %E Felix Berkenkamp %F pmlr-v235-jinnai24a %I PMLR %P 22326--22347 %U https://proceedings.mlr.press/v235/jinnai24a.html %V 235 %X Minimum Bayes Risk (MBR) decoding has been shown to be a powerful alternative to beam search decoding in a variety of text generation tasks. MBR decoding selects a hypothesis from a pool of hypotheses that has the least expected risk under a probability model according to a given utility function. Since it is impractical to compute the expected risk exactly over all possible hypotheses, two approximations are commonly used in MBR. First, it integrates over a sampled set of hypotheses rather than over all possible hypotheses. Second, it estimates the probability of each hypothesis using a Monte Carlo estimator. While the first approximation is necessary to make it computationally feasible, the second is not essential since we typically have access to the model probability at inference time. We propose model-based MBR (MBMBR), a variant of MBR that uses the model probability itself as the estimate of the probability distribution instead of the Monte Carlo estimate. We show analytically and empirically that the model-based estimate is more promising than the Monte Carlo estimate in text generation tasks. Our experiments show that MBMBR outperforms MBR in several text generation tasks, both with encoder-decoder models and with language models.
APA
Jinnai, Y., Morimura, T., Honda, U., Ariu, K. & Abe, K.. (2024). Model-Based Minimum Bayes Risk Decoding for Text Generation. Proceedings of the 41st International Conference on Machine Learning, in Proceedings of Machine Learning Research 235:22326-22347 Available from https://proceedings.mlr.press/v235/jinnai24a.html.

Related Material