@inproceedings{liu-etal-2026-timeprune,
title = "{T}ime{P}rune: Budgeted Temporal Pruning for Efficient Multimodal Time Series Analysis",
author = "Liu, Chenghao and
Liu, Qian and
Mahanti, Aniket and
Han, Jindong and
Li, Min and
Fan, Wei",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.805/",
pages = "18750--18771",
ISBN = "979-8-89176-459-0",
abstract = "The recent advancement of time series has witnessed Vision Language Models (VLMs) as a powerful tool for a variety of tasks of multimodal time series analysis, including forecasting, classification, question answering, etc. However, with time series as an explicit modality, inputs with long time series horizons yield excessive numeric tokens for VLMs, resulting in high inference latency. Moreover, such long contexts often contain redundant or noisy segments that might distract the model and thus degrade downstream performance. Although previous token pruning methods can reduce visual or textual tokens, they are mostly unimodal and do not explicitly explore multimodal semantics for time series. To address this, we propose \textit{TimePrune}, a plug-and-play temporal pruning framework for VLMs in multimodal time series analysis, which can accelerate VLM inference without fine-tuning the VLM backbone. Specifically, TimePrune formulates time series pruning as a budgeted temporal selection problem and introduces a Trimodal Budgeted Network to score temporal patches via jointly conditioning on linguistic and visual semantics. Then, under a fixed budget, this framework is optimized via the proposed Narrative-Calibrated Binned Grounding constraints, to further facilitate context-aware and temporally-coherent pruning. Extensive experiments in four time series tasks across application domains on multiple VLM backbones show that TimePrune could consistently achieve better performances. For example, on Qwen3VL-8B, TimePrune improves average forecasting performance by 33.91{\%}."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="liu-etal-2026-timeprune">
<titleInfo>
<title>TimePrune: Budgeted Temporal Pruning for Efficient Multimodal Time Series Analysis</title>
</titleInfo>
<name type="personal">
<namePart type="given">Chenghao</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qian</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aniket</namePart>
<namePart type="family">Mahanti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jindong</namePart>
<namePart type="family">Han</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Min</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Fan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-10</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hung-Yi</namePart>
<namePart type="family">Lee</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sunipa</namePart>
<namePart type="family">Dev</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jessy</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Desmond</namePart>
<namePart type="family">Elliott</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Budapest, Hungary</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>The recent advancement of time series has witnessed Vision Language Models (VLMs) as a powerful tool for a variety of tasks of multimodal time series analysis, including forecasting, classification, question answering, etc. However, with time series as an explicit modality, inputs with long time series horizons yield excessive numeric tokens for VLMs, resulting in high inference latency. Moreover, such long contexts often contain redundant or noisy segments that might distract the model and thus degrade downstream performance. Although previous token pruning methods can reduce visual or textual tokens, they are mostly unimodal and do not explicitly explore multimodal semantics for time series. To address this, we propose TimePrune, a plug-and-play temporal pruning framework for VLMs in multimodal time series analysis, which can accelerate VLM inference without fine-tuning the VLM backbone. Specifically, TimePrune formulates time series pruning as a budgeted temporal selection problem and introduces a Trimodal Budgeted Network to score temporal patches via jointly conditioning on linguistic and visual semantics. Then, under a fixed budget, this framework is optimized via the proposed Narrative-Calibrated Binned Grounding constraints, to further facilitate context-aware and temporally-coherent pruning. Extensive experiments in four time series tasks across application domains on multiple VLM backbones show that TimePrune could consistently achieve better performances. For example, on Qwen3VL-8B, TimePrune improves average forecasting performance by 33.91%.</abstract>
<identifier type="citekey">liu-etal-2026-timeprune</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.805/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>18750</start>
<end>18771</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T TimePrune: Budgeted Temporal Pruning for Efficient Multimodal Time Series Analysis
%A Liu, Chenghao
%A Liu, Qian
%A Mahanti, Aniket
%A Han, Jindong
%A Li, Min
%A Fan, Wei
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F liu-etal-2026-timeprune
%X The recent advancement of time series has witnessed Vision Language Models (VLMs) as a powerful tool for a variety of tasks of multimodal time series analysis, including forecasting, classification, question answering, etc. However, with time series as an explicit modality, inputs with long time series horizons yield excessive numeric tokens for VLMs, resulting in high inference latency. Moreover, such long contexts often contain redundant or noisy segments that might distract the model and thus degrade downstream performance. Although previous token pruning methods can reduce visual or textual tokens, they are mostly unimodal and do not explicitly explore multimodal semantics for time series. To address this, we propose TimePrune, a plug-and-play temporal pruning framework for VLMs in multimodal time series analysis, which can accelerate VLM inference without fine-tuning the VLM backbone. Specifically, TimePrune formulates time series pruning as a budgeted temporal selection problem and introduces a Trimodal Budgeted Network to score temporal patches via jointly conditioning on linguistic and visual semantics. Then, under a fixed budget, this framework is optimized via the proposed Narrative-Calibrated Binned Grounding constraints, to further facilitate context-aware and temporally-coherent pruning. Extensive experiments in four time series tasks across application domains on multiple VLM backbones show that TimePrune could consistently achieve better performances. For example, on Qwen3VL-8B, TimePrune improves average forecasting performance by 33.91%.
%U https://aclanthology.org/2026.emnlp-main.805/
%P 18750-18771
Markdown (Informal)
[TimePrune: Budgeted Temporal Pruning for Efficient Multimodal Time Series Analysis](https://aclanthology.org/2026.emnlp-main.805/) (Liu et al., EMNLP 2026)
ACL