@inproceedings{zhang-etal-2026-ect,
title = "{ECT}-{CLIP}: Resource-Efficient {CT} Vision-Language Pretraining via Negative Scaling",
author = "Zhang, Hanwen and
Liu, Yao and
Yang, Jiaye and
Dai, Die and
Liu, Qiao and
Xie, Yutong and
Wang, Peng",
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.991/",
pages = "22822--22834",
ISBN = "979-8-89176-459-0",
abstract = "Vision-Language Pretraining (VLP) for 3D CT scans is severely constrained by the massive GPU memory required for high-resolution volumes and large contrastive batches. We propose \textbf{ECT-CLIP}, a resource-efficient CT VLP framework that achieves negative scaling{---}scaling up both the quantity and quality of negative samples{---}under limited computational resources. To scale the quantity, we exploit the memory asymmetry between CT and text encoding to decouple the image and text batch sizes, thereby enlarging the negative pool. To scale the quality, we design two complementary strategies: (1) deterministically separating false negatives from informative hard negatives via UMLS concept normalization, and (2) fusing multiple frozen medical text encoders to enrich semantic diversity without extra training cost. Trained on a single consumer-grade RTX 4090 GPU, ECT-CLIP achieves state-of-the-art zero-shot classification AUCs of 83.4{\%} on CT-RATE and 75.0{\%} on the external RAD-ChestCT benchmark, outperforming prior methods that require 4{--}8 high-performance GPUs. Extensive ablation studies validate the effectiveness of each negative-scaling component."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="zhang-etal-2026-ect">
<titleInfo>
<title>ECT-CLIP: Resource-Efficient CT Vision-Language Pretraining via Negative Scaling</title>
</titleInfo>
<name type="personal">
<namePart type="given">Hanwen</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yao</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiaye</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Die</namePart>
<namePart type="family">Dai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Qiao</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yutong</namePart>
<namePart type="family">Xie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Peng</namePart>
<namePart type="family">Wang</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>Vision-Language Pretraining (VLP) for 3D CT scans is severely constrained by the massive GPU memory required for high-resolution volumes and large contrastive batches. We propose ECT-CLIP, a resource-efficient CT VLP framework that achieves negative scaling—scaling up both the quantity and quality of negative samples—under limited computational resources. To scale the quantity, we exploit the memory asymmetry between CT and text encoding to decouple the image and text batch sizes, thereby enlarging the negative pool. To scale the quality, we design two complementary strategies: (1) deterministically separating false negatives from informative hard negatives via UMLS concept normalization, and (2) fusing multiple frozen medical text encoders to enrich semantic diversity without extra training cost. Trained on a single consumer-grade RTX 4090 GPU, ECT-CLIP achieves state-of-the-art zero-shot classification AUCs of 83.4% on CT-RATE and 75.0% on the external RAD-ChestCT benchmark, outperforming prior methods that require 4–8 high-performance GPUs. Extensive ablation studies validate the effectiveness of each negative-scaling component.</abstract>
<identifier type="citekey">zhang-etal-2026-ect</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.991/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>22822</start>
<end>22834</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T ECT-CLIP: Resource-Efficient CT Vision-Language Pretraining via Negative Scaling
%A Zhang, Hanwen
%A Liu, Yao
%A Yang, Jiaye
%A Dai, Die
%A Liu, Qiao
%A Xie, Yutong
%A Wang, Peng
%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 zhang-etal-2026-ect
%X Vision-Language Pretraining (VLP) for 3D CT scans is severely constrained by the massive GPU memory required for high-resolution volumes and large contrastive batches. We propose ECT-CLIP, a resource-efficient CT VLP framework that achieves negative scaling—scaling up both the quantity and quality of negative samples—under limited computational resources. To scale the quantity, we exploit the memory asymmetry between CT and text encoding to decouple the image and text batch sizes, thereby enlarging the negative pool. To scale the quality, we design two complementary strategies: (1) deterministically separating false negatives from informative hard negatives via UMLS concept normalization, and (2) fusing multiple frozen medical text encoders to enrich semantic diversity without extra training cost. Trained on a single consumer-grade RTX 4090 GPU, ECT-CLIP achieves state-of-the-art zero-shot classification AUCs of 83.4% on CT-RATE and 75.0% on the external RAD-ChestCT benchmark, outperforming prior methods that require 4–8 high-performance GPUs. Extensive ablation studies validate the effectiveness of each negative-scaling component.
%U https://aclanthology.org/2026.emnlp-main.991/
%P 22822-22834
Markdown (Informal)
[ECT-CLIP: Resource-Efficient CT Vision-Language Pretraining via Negative Scaling](https://aclanthology.org/2026.emnlp-main.991/) (Zhang et al., EMNLP 2026)
ACL