@inproceedings{dey-etal-2026-soft,
title = "Do Soft Prompts Have Useful Wavelet Structure? Negative Evidence In A Controlled Setup",
author = "Dey, Soumyadeep and
Tenneti, Srikanth and
Mishra, Bamdev and
Niranjan, Shobhit and
Datar, Mayur",
editor = "Sedoc, Jo{\~a}o and
Shu, Raphael and
Drozd, Aleksandr and
Akula, Arjun and
Rogers, Anna and
Giorgi, Salvatore and
Benton, Adrian",
booktitle = "Proceedings of the Seventh Workshop on Insights from Negative Results in {NLP}",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.insights-1.6/",
pages = "48--59",
ISBN = "979-8-89176-480-4",
abstract = "Soft prompts are learned embedding sequences, but nothing guarantees that their token axis behaves like a signal with useful coarse and fine structure. We test that premise with the Discrete Wavelet Transform (DWT), which rewrites a sequence into approximation (coarse, low-frequency) and detail (local, high-frequency) coefficients. Wavelets help images and audio when most squared coefficient mass concentrates in the approximation band and the detail coefficients become sparse. Trained soft prompts meet neither condition. Our primary test decomposes a converged soft prompt into DWT bands, fine-tunes one band or both, and compares against a matched-budget baseline that continues vanilla training with no DWT. A constructed positive control first confirms that the procedure recovers a target displacement when the displacement is confined to the trainable band and rejects it when the required band is frozen. On SST-2 (67K) and AG News (120K) with Haar wavelets, and on an SST-2 check with Daubechies-2, no band-selective mode reliably beats the baseline, and the two bands are interchangeable within 0.2{\textasciitilde}pts. Direct measurement of the converged prompt finds coefficient mass roughly balanced across the bands and detail coefficients far denser than classical wavelet theory expects. Related variants behave the same way. An end-to-end wavelet prompt tuner does not reliably outperform vanilla, and a token-diversity variant makes prompt tokens more diverse without improving accuracy. On 250 held-out FLAN generation examples, existing Wavelet KD and Standard KD checkpoints are indistinguishable under token F1 and ROUGE-L."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="dey-etal-2026-soft">
<titleInfo>
<title>Do Soft Prompts Have Useful Wavelet Structure? Negative Evidence In A Controlled Setup</title>
</titleInfo>
<name type="personal">
<namePart type="given">Soumyadeep</namePart>
<namePart type="family">Dey</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Srikanth</namePart>
<namePart type="family">Tenneti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Bamdev</namePart>
<namePart type="family">Mishra</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shobhit</namePart>
<namePart type="family">Niranjan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Mayur</namePart>
<namePart type="family">Datar</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 Seventh Workshop on Insights from Negative Results in NLP</title>
</titleInfo>
<name type="personal">
<namePart type="given">João</namePart>
<namePart type="family">Sedoc</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Raphael</namePart>
<namePart type="family">Shu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Aleksandr</namePart>
<namePart type="family">Drozd</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Arjun</namePart>
<namePart type="family">Akula</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Anna</namePart>
<namePart type="family">Rogers</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Salvatore</namePart>
<namePart type="family">Giorgi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Adrian</namePart>
<namePart type="family">Benton</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">979-8-89176-480-4</identifier>
</relatedItem>
<abstract>Soft prompts are learned embedding sequences, but nothing guarantees that their token axis behaves like a signal with useful coarse and fine structure. We test that premise with the Discrete Wavelet Transform (DWT), which rewrites a sequence into approximation (coarse, low-frequency) and detail (local, high-frequency) coefficients. Wavelets help images and audio when most squared coefficient mass concentrates in the approximation band and the detail coefficients become sparse. Trained soft prompts meet neither condition. Our primary test decomposes a converged soft prompt into DWT bands, fine-tunes one band or both, and compares against a matched-budget baseline that continues vanilla training with no DWT. A constructed positive control first confirms that the procedure recovers a target displacement when the displacement is confined to the trainable band and rejects it when the required band is frozen. On SST-2 (67K) and AG News (120K) with Haar wavelets, and on an SST-2 check with Daubechies-2, no band-selective mode reliably beats the baseline, and the two bands are interchangeable within 0.2~pts. Direct measurement of the converged prompt finds coefficient mass roughly balanced across the bands and detail coefficients far denser than classical wavelet theory expects. Related variants behave the same way. An end-to-end wavelet prompt tuner does not reliably outperform vanilla, and a token-diversity variant makes prompt tokens more diverse without improving accuracy. On 250 held-out FLAN generation examples, existing Wavelet KD and Standard KD checkpoints are indistinguishable under token F1 and ROUGE-L.</abstract>
<identifier type="citekey">dey-etal-2026-soft</identifier>
<location>
<url>https://aclanthology.org/2026.insights-1.6/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>48</start>
<end>59</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Do Soft Prompts Have Useful Wavelet Structure? Negative Evidence In A Controlled Setup
%A Dey, Soumyadeep
%A Tenneti, Srikanth
%A Mishra, Bamdev
%A Niranjan, Shobhit
%A Datar, Mayur
%Y Sedoc, João
%Y Shu, Raphael
%Y Drozd, Aleksandr
%Y Akula, Arjun
%Y Rogers, Anna
%Y Giorgi, Salvatore
%Y Benton, Adrian
%S Proceedings of the Seventh Workshop on Insights from Negative Results in NLP
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-480-4
%F dey-etal-2026-soft
%X Soft prompts are learned embedding sequences, but nothing guarantees that their token axis behaves like a signal with useful coarse and fine structure. We test that premise with the Discrete Wavelet Transform (DWT), which rewrites a sequence into approximation (coarse, low-frequency) and detail (local, high-frequency) coefficients. Wavelets help images and audio when most squared coefficient mass concentrates in the approximation band and the detail coefficients become sparse. Trained soft prompts meet neither condition. Our primary test decomposes a converged soft prompt into DWT bands, fine-tunes one band or both, and compares against a matched-budget baseline that continues vanilla training with no DWT. A constructed positive control first confirms that the procedure recovers a target displacement when the displacement is confined to the trainable band and rejects it when the required band is frozen. On SST-2 (67K) and AG News (120K) with Haar wavelets, and on an SST-2 check with Daubechies-2, no band-selective mode reliably beats the baseline, and the two bands are interchangeable within 0.2~pts. Direct measurement of the converged prompt finds coefficient mass roughly balanced across the bands and detail coefficients far denser than classical wavelet theory expects. Related variants behave the same way. An end-to-end wavelet prompt tuner does not reliably outperform vanilla, and a token-diversity variant makes prompt tokens more diverse without improving accuracy. On 250 held-out FLAN generation examples, existing Wavelet KD and Standard KD checkpoints are indistinguishable under token F1 and ROUGE-L.
%U https://aclanthology.org/2026.insights-1.6/
%P 48-59
Markdown (Informal)
[Do Soft Prompts Have Useful Wavelet Structure? Negative Evidence In A Controlled Setup](https://aclanthology.org/2026.insights-1.6/) (Dey et al., insights 2026)
ACL