<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="lim-lauw-2023-disentangling">
    <titleInfo>
        <title>Disentangling Transformer Language Models as Superposed Topic Models</title>
    </titleInfo>
    <name type="personal">
        <namePart type="given">Jia</namePart>
        <namePart type="given">Peng</namePart>
        <namePart type="family">Lim</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <name type="personal">
        <namePart type="given">Hady</namePart>
        <namePart type="family">Lauw</namePart>
        <role>
            <roleTerm authority="marcrelator" type="text">author</roleTerm>
        </role>
    </name>
    <originInfo>
        <dateIssued>2023-12</dateIssued>
    </originInfo>
    <typeOfResource>text</typeOfResource>
    <relatedItem type="host">
        <titleInfo>
            <title>Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing</title>
        </titleInfo>
        <name type="personal">
            <namePart type="given">Houda</namePart>
            <namePart type="family">Bouamor</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Juan</namePart>
            <namePart type="family">Pino</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <name type="personal">
            <namePart type="given">Kalika</namePart>
            <namePart type="family">Bali</namePart>
            <role>
                <roleTerm authority="marcrelator" type="text">editor</roleTerm>
            </role>
        </name>
        <originInfo>
            <publisher>Association for Computational Linguistics</publisher>
            <place>
                <placeTerm type="text">Singapore</placeTerm>
            </place>
        </originInfo>
        <genre authority="marcgt">conference publication</genre>
    </relatedItem>
    <abstract>Topic Modelling is an established research area where the quality of a given topic is measured using coherence metrics. Often, we infer topics from Neural Topic Models (NTM) by interpreting their decoder weights, consisting of top-activated words projected from individual neurons. Transformer-based Language Models (TLM) similarly consist of decoder weights. However, due to its hypothesised superposition properties, the final logits originating from the residual path are considered uninterpretable. Therefore, we posit that we can interpret TLM as superposed NTM by proposing a novel weight-based, model-agnostic and corpus-agnostic approach to search and disentangle decoder-only TLM, potentially mapping individual neurons to multiple coherent topics. Our results show that it is empirically feasible to disentangle coherent topics from GPT-2 models using the Wikipedia corpus. We validate this approach for GPT-2 models using Zero-Shot Topic Modelling. Finally, we extend the proposed approach to disentangle and analyse LLaMA models.</abstract>
    <identifier type="citekey">lim-lauw-2023-disentangling</identifier>
    <identifier type="doi">10.18653/v1/2023.emnlp-main.534</identifier>
    <location>
        <url>https://aclanthology.org/2023.emnlp-main.534</url>
    </location>
    <part>
        <date>2023-12</date>
        <extent unit="page">
            <start>8646</start>
            <end>8666</end>
        </extent>
    </part>
</mods>
</modsCollection>

%0 Conference Proceedings
%T Disentangling Transformer Language Models as Superposed Topic Models
%A Lim, Jia Peng
%A Lauw, Hady
%Y Bouamor, Houda
%Y Pino, Juan
%Y Bali, Kalika
%S Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing
%D 2023
%8 December
%I Association for Computational Linguistics
%C Singapore
%F lim-lauw-2023-disentangling
%X Topic Modelling is an established research area where the quality of a given topic is measured using coherence metrics. Often, we infer topics from Neural Topic Models (NTM) by interpreting their decoder weights, consisting of top-activated words projected from individual neurons. Transformer-based Language Models (TLM) similarly consist of decoder weights. However, due to its hypothesised superposition properties, the final logits originating from the residual path are considered uninterpretable. Therefore, we posit that we can interpret TLM as superposed NTM by proposing a novel weight-based, model-agnostic and corpus-agnostic approach to search and disentangle decoder-only TLM, potentially mapping individual neurons to multiple coherent topics. Our results show that it is empirically feasible to disentangle coherent topics from GPT-2 models using the Wikipedia corpus. We validate this approach for GPT-2 models using Zero-Shot Topic Modelling. Finally, we extend the proposed approach to disentangle and analyse LLaMA models.
%R 10.18653/v1/2023.emnlp-main.534
%U https://aclanthology.org/2023.emnlp-main.534
%U https://doi.org/10.18653/v1/2023.emnlp-main.534
%P 8646-8666

Markdown (Informal)

[Disentangling Transformer Language Models as Superposed Topic Models](https://aclanthology.org/2023.emnlp-main.534) (Lim & Lauw, EMNLP 2023)

• Disentangling Transformer Language Models as Superposed Topic Models (Lim & Lauw, EMNLP 2023)

ACL

• Jia Peng Lim and Hady Lauw. 2023. Disentangling Transformer Language Models as Superposed Topic Models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, pages 8646–8666, Singapore. Association for Computational Linguistics.