Inferring from Logits: Exploring Best Practices for Decoding-Free Generative Candidate Selection

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ma, Mingyu Derek, Ding, Yanna, Huang, Zijie, Gao, Jianxi, Sun, Yizhou, Wang, Wei
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866912208564256768
author Ma, Mingyu Derek
Ding, Yanna
Huang, Zijie
Gao, Jianxi
Sun, Yizhou
Wang, Wei
author_facet Ma, Mingyu Derek
Ding, Yanna
Huang, Zijie
Gao, Jianxi
Sun, Yizhou
Wang, Wei
contents Generative Language Models rely on autoregressive decoding to produce the output sequence token by token. Many tasks such as preference optimization, require the model to produce task-level output consisting of multiple tokens directly by selecting candidates from a pool as predictions. Determining a task-level prediction from candidates using the ordinary token-level decoding mechanism is constrained by time-consuming decoding and interrupted gradients by discrete token selection. Existing works have been using decoding-free candidate selection methods to obtain candidate probability from initial output logits over vocabulary. Though these estimation methods are widely used, they are not systematically evaluated, especially on end tasks. We introduce an evaluation of a comprehensive collection of decoding-free candidate selection approaches on a comprehensive set of tasks, including five multiple-choice QA tasks with a small candidate pool and four clinical decision tasks with a massive amount of candidates, some with 10k+ options. We evaluate the estimation methods paired with a wide spectrum of foundation LMs covering different architectures, sizes and training paradigms. The results and insights from our analysis inform the future model design.
format Preprint
id arxiv_https___arxiv_org_abs_2501_17338
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Inferring from Logits: Exploring Best Practices for Decoding-Free Generative Candidate Selection
Ma, Mingyu Derek
Ding, Yanna
Huang, Zijie
Gao, Jianxi
Sun, Yizhou
Wang, Wei
Computation and Language
Artificial Intelligence
Machine Learning
Generative Language Models rely on autoregressive decoding to produce the output sequence token by token. Many tasks such as preference optimization, require the model to produce task-level output consisting of multiple tokens directly by selecting candidates from a pool as predictions. Determining a task-level prediction from candidates using the ordinary token-level decoding mechanism is constrained by time-consuming decoding and interrupted gradients by discrete token selection. Existing works have been using decoding-free candidate selection methods to obtain candidate probability from initial output logits over vocabulary. Though these estimation methods are widely used, they are not systematically evaluated, especially on end tasks. We introduce an evaluation of a comprehensive collection of decoding-free candidate selection approaches on a comprehensive set of tasks, including five multiple-choice QA tasks with a small candidate pool and four clinical decision tasks with a massive amount of candidates, some with 10k+ options. We evaluate the estimation methods paired with a wide spectrum of foundation LMs covering different architectures, sizes and training paradigms. The results and insights from our analysis inform the future model design.
title Inferring from Logits: Exploring Best Practices for Decoding-Free Generative Candidate Selection
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2501.17338