@inproceedings{xie-etal-2025-word,
title = "Word Salad Chopper: Reasoning Models Waste A Ton Of Decoding Budget On Useless Repetitions, Self-Knowingly",
author = "Xie, Wenya and
Zhong, Shaochen and
Le, Hoang Anh Duy and
Xu, Zhaozhuo and
Xie, Jianwen and
Liu, Zirui",
editor = "Christodoulopoulos, Christos and
Chakraborty, Tanmoy and
Rose, Carolyn and
Peng, Violet",
booktitle = "Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing",
month = nov,
year = "2025",
address = "Suzhou, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2025.emnlp-main.1705/",
doi = "10.18653/v1/2025.emnlp-main.1705",
pages = "33588--33598",
ISBN = "979-8-89176-332-6",
abstract = "Large Reasoning Models (LRMs) are often bottlenecked by the high cost of output tokens. We show that a significant portion of these tokens are useless self-repetitions {---} what we call ``word salad'' {---} that exhaust the decoding budget without adding value. Interestingly, we observe that LRMs are self-aware when trapped in these loops: the hidden states of `{`} tokens trailing each reasoning chunk exhibit patterns that allow us to detect word salad behavior on-the-fly via a single linear classifier. Once detected, a simple chop appended by a straightforward regeneration prompt yields substantial length savings with minimal quality loss. Our work offers WordSaladChopper (WSC) {---} a lightweight, turnkey component for LRM that is minimally invasive to its reasoning trajectory. Given its low overhead, strong savings, and the lack of semantic value of word salad tokens, we believe it is not too far-fetched to argue that WSC {---} or a similar component {---} is a must-have for all LRM applications with user experience in mind."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="xie-etal-2025-word">
<titleInfo>
<title>Word Salad Chopper: Reasoning Models Waste A Ton Of Decoding Budget On Useless Repetitions, Self-Knowingly</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wenya</namePart>
<namePart type="family">Xie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shaochen</namePart>
<namePart type="family">Zhong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hoang</namePart>
<namePart type="given">Anh</namePart>
<namePart type="given">Duy</namePart>
<namePart type="family">Le</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhaozhuo</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jianwen</namePart>
<namePart type="family">Xie</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zirui</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2025-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing</title>
</titleInfo>
<name type="personal">
<namePart type="given">Christos</namePart>
<namePart type="family">Christodoulopoulos</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tanmoy</namePart>
<namePart type="family">Chakraborty</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Carolyn</namePart>
<namePart type="family">Rose</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Violet</namePart>
<namePart type="family">Peng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Suzhou, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-332-6</identifier>
</relatedItem>
<abstract>Large Reasoning Models (LRMs) are often bottlenecked by the high cost of output tokens. We show that a significant portion of these tokens are useless self-repetitions — what we call “word salad” — that exhaust the decoding budget without adding value. Interestingly, we observe that LRMs are self-aware when trapped in these loops: the hidden states of “ tokens trailing each reasoning chunk exhibit patterns that allow us to detect word salad behavior on-the-fly via a single linear classifier. Once detected, a simple chop appended by a straightforward regeneration prompt yields substantial length savings with minimal quality loss. Our work offers WordSaladChopper (WSC) — a lightweight, turnkey component for LRM that is minimally invasive to its reasoning trajectory. Given its low overhead, strong savings, and the lack of semantic value of word salad tokens, we believe it is not too far-fetched to argue that WSC — or a similar component — is a must-have for all LRM applications with user experience in mind.</abstract>
<identifier type="citekey">xie-etal-2025-word</identifier>
<identifier type="doi">10.18653/v1/2025.emnlp-main.1705</identifier>
<location>
<url>https://aclanthology.org/2025.emnlp-main.1705/</url>
</location>
<part>
<date>2025-11</date>
<extent unit="page">
<start>33588</start>
<end>33598</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Word Salad Chopper: Reasoning Models Waste A Ton Of Decoding Budget On Useless Repetitions, Self-Knowingly
%A Xie, Wenya
%A Zhong, Shaochen
%A Le, Hoang Anh Duy
%A Xu, Zhaozhuo
%A Xie, Jianwen
%A Liu, Zirui
%Y Christodoulopoulos, Christos
%Y Chakraborty, Tanmoy
%Y Rose, Carolyn
%Y Peng, Violet
%S Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing
%D 2025
%8 November
%I Association for Computational Linguistics
%C Suzhou, China
%@ 979-8-89176-332-6
%F xie-etal-2025-word
%X Large Reasoning Models (LRMs) are often bottlenecked by the high cost of output tokens. We show that a significant portion of these tokens are useless self-repetitions — what we call “word salad” — that exhaust the decoding budget without adding value. Interestingly, we observe that LRMs are self-aware when trapped in these loops: the hidden states of “ tokens trailing each reasoning chunk exhibit patterns that allow us to detect word salad behavior on-the-fly via a single linear classifier. Once detected, a simple chop appended by a straightforward regeneration prompt yields substantial length savings with minimal quality loss. Our work offers WordSaladChopper (WSC) — a lightweight, turnkey component for LRM that is minimally invasive to its reasoning trajectory. Given its low overhead, strong savings, and the lack of semantic value of word salad tokens, we believe it is not too far-fetched to argue that WSC — or a similar component — is a must-have for all LRM applications with user experience in mind.
%R 10.18653/v1/2025.emnlp-main.1705
%U https://aclanthology.org/2025.emnlp-main.1705/
%U https://doi.org/10.18653/v1/2025.emnlp-main.1705
%P 33588-33598
Markdown (Informal)
[Word Salad Chopper: Reasoning Models Waste A Ton Of Decoding Budget On Useless Repetitions, Self-Knowingly](https://aclanthology.org/2025.emnlp-main.1705/) (Xie et al., EMNLP 2025)
ACL