Advancing Sequential Numerical Prediction in Autoregressive Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Fei, Xiang, Lu, Jinghui, Sun, Qi, Feng, Hao, Wang, Yanjie, Shi, Wei, Wang, An-Lan, Tang, Jingqun, Huang, Can
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916763878293504
author Fei, Xiang
Lu, Jinghui
Sun, Qi
Feng, Hao
Wang, Yanjie
Shi, Wei
Wang, An-Lan
Tang, Jingqun
Huang, Can
author_facet Fei, Xiang
Lu, Jinghui
Sun, Qi
Feng, Hao
Wang, Yanjie
Shi, Wei
Wang, An-Lan
Tang, Jingqun
Huang, Can
contents Autoregressive models have become the de facto choice for sequence generation tasks, but standard approaches treat digits as independent tokens and apply cross-entropy loss, overlooking the coherent structure of numerical sequences. This paper introduces Numerical Token Integrity Loss (NTIL) to address this gap. NTIL operates at two levels: (1) token-level, where it extends the Earth Mover's Distance (EMD) to preserve ordinal relationships between numerical values, and (2) sequence-level, where it penalizes the overall discrepancy between the predicted and actual sequences. This dual approach improves numerical prediction and integrates effectively with LLMs/MLLMs. Extensive experiments show significant performance improvements with NTIL.
format Preprint
id arxiv_https___arxiv_org_abs_2505_13077
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Advancing Sequential Numerical Prediction in Autoregressive Models
Fei, Xiang
Lu, Jinghui
Sun, Qi
Feng, Hao
Wang, Yanjie
Shi, Wei
Wang, An-Lan
Tang, Jingqun
Huang, Can
Computation and Language
Artificial Intelligence
Machine Learning
Autoregressive models have become the de facto choice for sequence generation tasks, but standard approaches treat digits as independent tokens and apply cross-entropy loss, overlooking the coherent structure of numerical sequences. This paper introduces Numerical Token Integrity Loss (NTIL) to address this gap. NTIL operates at two levels: (1) token-level, where it extends the Earth Mover's Distance (EMD) to preserve ordinal relationships between numerical values, and (2) sequence-level, where it penalizes the overall discrepancy between the predicted and actual sequences. This dual approach improves numerical prediction and integrates effectively with LLMs/MLLMs. Extensive experiments show significant performance improvements with NTIL.
title Advancing Sequential Numerical Prediction in Autoregressive Models
topic Computation and Language
Artificial Intelligence
Machine Learning
url https://arxiv.org/abs/2505.13077