@inproceedings{wang-etal-2024-context-former,
title = "In-Context Former: Lightning-fast Compressing Context for Large Language Model",
author = "Wang, Xiangfeng and
Chen, Zaiyi and
Xu, Tong and
Xie, Zheyong and
He, Yongyi and
Chen, Enhong",
editor = "Al-Onaizan, Yaser and
Bansal, Mohit and
Chen, Yun-Nung",
booktitle = "Findings of the Association for Computational Linguistics: EMNLP 2024",
month = nov,
year = "2024",
address = "Miami, Florida, USA",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2024.findings-emnlp.138",
pages = "2445--2460",
abstract = "With the rising popularity of Transformer-based large language models (LLMs), reducing their high inference costs has become a significant research focus. One effective approach to mitigate these costs is compressing the long input contexts. Existing methods typically leverage the self-attention mechanism of the large model itself for context compression. While these methods have achieved notable results, the compression process still entails quadratic complexity. To mitigate this limitation, we propose the In-Context Former (IC-Former). This method does not rely on the target large model but instead utilizes cross-attention mechanisms to extract and condense information from the contextual embeddings. The computational overhead of our method grows linearly with the compression range. Experimental results indicate that our method requires only 1/32 of the floating-point operations of the baseline during compression and improves processing speed by 68 to 112 times while achieving 90{\%} of the baseline performance on evaluation metrics. Additionally, IC-Former demonstrates strong regularity in its interactions with the context, enhancing its interpretability. Overall, IC-Former significantly reduces compression costs, making real-time compression scenarios feasible.",
}
<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2024-context-former">
<titleInfo>
<title>In-Context Former: Lightning-fast Compressing Context for Large Language Model</title>
</titleInfo>
<name type="personal">
<namePart type="given">Xiangfeng</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zaiyi</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zheyong</namePart>
<namePart type="family">Xie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yongyi</namePart>
<namePart type="family">He</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Enhong</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2024-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: EMNLP 2024</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yaser</namePart>
<namePart type="family">Al-Onaizan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mohit</namePart>
<namePart type="family">Bansal</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yun-Nung</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Miami, Florida, USA</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
</relatedItem>
<abstract>With the rising popularity of Transformer-based large language models (LLMs), reducing their high inference costs has become a significant research focus. One effective approach to mitigate these costs is compressing the long input contexts. Existing methods typically leverage the self-attention mechanism of the large model itself for context compression. While these methods have achieved notable results, the compression process still entails quadratic complexity. To mitigate this limitation, we propose the In-Context Former (IC-Former). This method does not rely on the target large model but instead utilizes cross-attention mechanisms to extract and condense information from the contextual embeddings. The computational overhead of our method grows linearly with the compression range. Experimental results indicate that our method requires only 1/32 of the floating-point operations of the baseline during compression and improves processing speed by 68 to 112 times while achieving 90% of the baseline performance on evaluation metrics. Additionally, IC-Former demonstrates strong regularity in its interactions with the context, enhancing its interpretability. Overall, IC-Former significantly reduces compression costs, making real-time compression scenarios feasible.</abstract>
<identifier type="citekey">wang-etal-2024-context-former</identifier>
<location>
<url>https://aclanthology.org/2024.findings-emnlp.138</url>
</location>
<part>
<date>2024-11</date>
<extent unit="page">
<start>2445</start>
<end>2460</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T In-Context Former: Lightning-fast Compressing Context for Large Language Model
%A Wang, Xiangfeng
%A Chen, Zaiyi
%A Xu, Tong
%A Xie, Zheyong
%A He, Yongyi
%A Chen, Enhong
%Y Al-Onaizan, Yaser
%Y Bansal, Mohit
%Y Chen, Yun-Nung
%S Findings of the Association for Computational Linguistics: EMNLP 2024
%D 2024
%8 November
%I Association for Computational Linguistics
%C Miami, Florida, USA
%F wang-etal-2024-context-former
%X With the rising popularity of Transformer-based large language models (LLMs), reducing their high inference costs has become a significant research focus. One effective approach to mitigate these costs is compressing the long input contexts. Existing methods typically leverage the self-attention mechanism of the large model itself for context compression. While these methods have achieved notable results, the compression process still entails quadratic complexity. To mitigate this limitation, we propose the In-Context Former (IC-Former). This method does not rely on the target large model but instead utilizes cross-attention mechanisms to extract and condense information from the contextual embeddings. The computational overhead of our method grows linearly with the compression range. Experimental results indicate that our method requires only 1/32 of the floating-point operations of the baseline during compression and improves processing speed by 68 to 112 times while achieving 90% of the baseline performance on evaluation metrics. Additionally, IC-Former demonstrates strong regularity in its interactions with the context, enhancing its interpretability. Overall, IC-Former significantly reduces compression costs, making real-time compression scenarios feasible.
%U https://aclanthology.org/2024.findings-emnlp.138
%P 2445-2460
Markdown (Informal)
[In-Context Former: Lightning-fast Compressing Context for Large Language Model](https://aclanthology.org/2024.findings-emnlp.138) (Wang et al., Findings 2024)
ACL