Regression Aware Inference with LLMs

Michal Lukasik, Harikrishna Narasimhan, Aditya Menon, Felix Yu, Sanjiv Kumar


Abstract
Large language models (LLMs) have shown strong results on a range of applications, including regression and scoring tasks.Typically, one obtains outputs from an LLM via autoregressive sampling from the model’s output distribution. We show that this inference strategy can be sub-optimal for common regression and scoring evaluation metrics. As a remedy, we build on prior work on Minimum Bayes Risk decoding,and propose alternate inference strategies that estimate the Bayes-optimal solution for regression and scoring metrics in closed-form from sampled responses.We show that our proposal significantly improves over baselines across datasets and models.
Anthology ID:
2024.findings-emnlp.799
Volume:
Findings of the Association for Computational Linguistics: EMNLP 2024
Month:
November
Year:
2024
Address:
Miami, Florida, USA
Editors:
Yaser Al-Onaizan, Mohit Bansal, Yun-Nung Chen
Venue:
Findings
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
13667–13678
Language:
URL:
https://aclanthology.org/2024.findings-emnlp.799
DOI:
Bibkey:
Cite (ACL):
Michal Lukasik, Harikrishna Narasimhan, Aditya Menon, Felix Yu, and Sanjiv Kumar. 2024. Regression Aware Inference with LLMs. In Findings of the Association for Computational Linguistics: EMNLP 2024, pages 13667–13678, Miami, Florida, USA. Association for Computational Linguistics.
Cite (Informal):
Regression Aware Inference with LLMs (Lukasik et al., Findings 2024)
Copy Citation:
PDF:
https://aclanthology.org/2024.findings-emnlp.799.pdf