@inproceedings{huang-etal-2026-nonsense,
title = "Nonsense Helps: Prompt Space Perturbation Broadens Reasoning Exploration",
author = "Huang, Langlin and
Huang, Chengsong and
Li, Jinyuan and
Cai, Donghong and
Yang, Yuyi and
Huang, Jiaxin",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {EMNLP} 2026",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-emnlp.553/",
pages = "11655--11672",
ISBN = "979-8-89176-462-0",
abstract = "Reinforcement learning with verifiable rewards (e.g., GRPO) significantly advances LLM reasoning. However, in complex tasks, GRPO frequently encounters the ``zero-advantage problem'': when all rollouts for a query fail, the relative advantage collapses to zero, wasting computational budget. While resampling with more budget is a common remedy, the static sampling policy inherently constrains reasoning exploration, limiting the success rate. In this paper, we propose Lorem Perturbation for Exploration (LoPE), a simple training framework to break this exploration bottleneck. Motivated by the intuition that task-irrelevant prompt perturbations can shift the output distribution, LoPE prepends sequences stochastically assembled from the Lorem Ipsum vocabulary to the prompts before resampling, forcing the model to explore broader reasoning pathways. Experiments across 1.7B, 4B, and 7B models show LoPE significantly outperforms resampling with original prompts. Our findings, including the effectiveness of other low-perplexity Latin-based random sequences, establish LoPE as a strong baseline for broadening LLM exploration."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2026-nonsense">
<titleInfo>
<title>Nonsense Helps: Prompt Space Perturbation Broadens Reasoning Exploration</title>
</titleInfo>
<name type="personal">
<namePart type="given">Langlin</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chengsong</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jinyuan</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Donghong</namePart>
<namePart type="family">Cai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuyi</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiaxin</namePart>
<namePart type="family">Huang</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>Findings of the Association for Computational Linguistics: EMNLP 2026</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">979-8-89176-462-0</identifier>
</relatedItem>
<abstract>Reinforcement learning with verifiable rewards (e.g., GRPO) significantly advances LLM reasoning. However, in complex tasks, GRPO frequently encounters the “zero-advantage problem”: when all rollouts for a query fail, the relative advantage collapses to zero, wasting computational budget. While resampling with more budget is a common remedy, the static sampling policy inherently constrains reasoning exploration, limiting the success rate. In this paper, we propose Lorem Perturbation for Exploration (LoPE), a simple training framework to break this exploration bottleneck. Motivated by the intuition that task-irrelevant prompt perturbations can shift the output distribution, LoPE prepends sequences stochastically assembled from the Lorem Ipsum vocabulary to the prompts before resampling, forcing the model to explore broader reasoning pathways. Experiments across 1.7B, 4B, and 7B models show LoPE significantly outperforms resampling with original prompts. Our findings, including the effectiveness of other low-perplexity Latin-based random sequences, establish LoPE as a strong baseline for broadening LLM exploration.</abstract>
<identifier type="citekey">huang-etal-2026-nonsense</identifier>
<location>
<url>https://aclanthology.org/2026.findings-emnlp.553/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>11655</start>
<end>11672</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Nonsense Helps: Prompt Space Perturbation Broadens Reasoning Exploration
%A Huang, Langlin
%A Huang, Chengsong
%A Li, Jinyuan
%A Cai, Donghong
%A Yang, Yuyi
%A Huang, Jiaxin
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Findings of the Association for Computational Linguistics: EMNLP 2026
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-462-0
%F huang-etal-2026-nonsense
%X Reinforcement learning with verifiable rewards (e.g., GRPO) significantly advances LLM reasoning. However, in complex tasks, GRPO frequently encounters the “zero-advantage problem”: when all rollouts for a query fail, the relative advantage collapses to zero, wasting computational budget. While resampling with more budget is a common remedy, the static sampling policy inherently constrains reasoning exploration, limiting the success rate. In this paper, we propose Lorem Perturbation for Exploration (LoPE), a simple training framework to break this exploration bottleneck. Motivated by the intuition that task-irrelevant prompt perturbations can shift the output distribution, LoPE prepends sequences stochastically assembled from the Lorem Ipsum vocabulary to the prompts before resampling, forcing the model to explore broader reasoning pathways. Experiments across 1.7B, 4B, and 7B models show LoPE significantly outperforms resampling with original prompts. Our findings, including the effectiveness of other low-perplexity Latin-based random sequences, establish LoPE as a strong baseline for broadening LLM exploration.
%U https://aclanthology.org/2026.findings-emnlp.553/
%P 11655-11672
Markdown (Informal)
[Nonsense Helps: Prompt Space Perturbation Broadens Reasoning Exploration](https://aclanthology.org/2026.findings-emnlp.553/) (Huang et al., Findings 2026)
ACL