@inproceedings{le-etal-2026-tokens,
title = "Not All Tokens Are Equally Important: Auditing and Compressing Redundancy in {LLM} Reasoning",
author = "Le, Nguyen-Khang and
Nguyen, Dieu-Hien and
Do, Truong Dinh and
Nguyen, Le-Minh",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the {A}sia-Pacific Chapter of the {A}ssociation for {C}omputational {L}inguistics (Volume 1: Long Papers)",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.aacl-long.268/",
pages = "5942--5962",
ISBN = "979-8-89176-439-2",
abstract = "Chain-of-thought traces drive strong LLM accuracy but dominate inference cost: output length sets both monetary cost and decoding latency. We ask which tokens in a reasoning trace actually carry the answer. A token-level analysis of the model{'}s own next-token distribution identifies three redundancies - discourse-element removal, tokenizer-aware abbreviation, and substitution of prose causal connectives with an arrow ``X -{\ensuremath{>}} Y''. All three leave the predictive distribution close to the original under Jensen-Shannon divergence and preserve the greedy next token in the vast majority of prefixes. We instantiate the analysis as CONCISON, a single training-free system prompt that operationalises the three rules and protects numeric, equation, and code tokens verbatim. Across 18 models from six families (2B-120B) and five reasoning benchmarks, CONCISON cuts output tokens by 24-62{\%} on strong models at -1.1pp average accuracy, outperforms token-budget prompting in a single decoding call, and matches trained routers on the accuracy-cost frontier without any training, model pool, or extra infrastructure."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="le-etal-2026-tokens">
<titleInfo>
<title>Not All Tokens Are Equally Important: Auditing and Compressing Redundancy in LLM Reasoning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Nguyen-Khang</namePart>
<namePart type="family">Le</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Dieu-Hien</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Truong</namePart>
<namePart type="given">Dinh</namePart>
<namePart type="family">Do</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Le-Minh</namePart>
<namePart type="family">Nguyen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-11</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yue</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ruifeng</namePart>
<namePart type="family">Xu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ayu</namePart>
<namePart type="family">Purwarianti</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tong</namePart>
<namePart type="family">Rong</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jun</namePart>
<namePart type="family">Suzuki</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">Hengqin, China</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-439-2</identifier>
</relatedItem>
<abstract>Chain-of-thought traces drive strong LLM accuracy but dominate inference cost: output length sets both monetary cost and decoding latency. We ask which tokens in a reasoning trace actually carry the answer. A token-level analysis of the model’s own next-token distribution identifies three redundancies - discourse-element removal, tokenizer-aware abbreviation, and substitution of prose causal connectives with an arrow “X -\ensuremath> Y”. All three leave the predictive distribution close to the original under Jensen-Shannon divergence and preserve the greedy next token in the vast majority of prefixes. We instantiate the analysis as CONCISON, a single training-free system prompt that operationalises the three rules and protects numeric, equation, and code tokens verbatim. Across 18 models from six families (2B-120B) and five reasoning benchmarks, CONCISON cuts output tokens by 24-62% on strong models at -1.1pp average accuracy, outperforms token-budget prompting in a single decoding call, and matches trained routers on the accuracy-cost frontier without any training, model pool, or extra infrastructure.</abstract>
<identifier type="citekey">le-etal-2026-tokens</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-long.268/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>5942</start>
<end>5962</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Not All Tokens Are Equally Important: Auditing and Compressing Redundancy in LLM Reasoning
%A Le, Nguyen-Khang
%A Nguyen, Dieu-Hien
%A Do, Truong Dinh
%A Nguyen, Le-Minh
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers)
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-439-2
%F le-etal-2026-tokens
%X Chain-of-thought traces drive strong LLM accuracy but dominate inference cost: output length sets both monetary cost and decoding latency. We ask which tokens in a reasoning trace actually carry the answer. A token-level analysis of the model’s own next-token distribution identifies three redundancies - discourse-element removal, tokenizer-aware abbreviation, and substitution of prose causal connectives with an arrow “X -\ensuremath> Y”. All three leave the predictive distribution close to the original under Jensen-Shannon divergence and preserve the greedy next token in the vast majority of prefixes. We instantiate the analysis as CONCISON, a single training-free system prompt that operationalises the three rules and protects numeric, equation, and code tokens verbatim. Across 18 models from six families (2B-120B) and five reasoning benchmarks, CONCISON cuts output tokens by 24-62% on strong models at -1.1pp average accuracy, outperforms token-budget prompting in a single decoding call, and matches trained routers on the accuracy-cost frontier without any training, model pool, or extra infrastructure.
%U https://aclanthology.org/2026.aacl-long.268/
%P 5942-5962
Markdown (Informal)
[Not All Tokens Are Equally Important: Auditing and Compressing Redundancy in LLM Reasoning](https://aclanthology.org/2026.aacl-long.268/) (Le et al., AACL-IJCNLP 2026)
ACL
- Nguyen-Khang Le, Dieu-Hien Nguyen, Truong Dinh Do, and Le-Minh Nguyen. 2026. Not All Tokens Are Equally Important: Auditing and Compressing Redundancy in LLM Reasoning. In Proceedings of the 15th International Joint Conference on Natural Language Processing and the 5th Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers), pages 5942–5962, Hengqin, China. Association for Computational Linguistics.