@inproceedings{le-etal-2026-unitor,
title = "{U}ni{T}o{R}: Towards Universal Token-level Routing for Efficient Reasoning in Large Language Models",
author = "Le, Nguyen-Khang and
Do, Truong Dinh and
Nguyen, Le-Minh",
editor = "Zhang, Yue and
Xu, Ruifeng and
Purwarianti, Ayu and
Rong, Tong and
Suzuki, Jun",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {IJCNLP}-{AACL} 2026",
month = nov,
year = "2026",
address = "Hengqin, China",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-aacl.212/",
pages = "4197--4215",
ISBN = "979-8-89176-444-6",
abstract = "Large Reasoning Models (LRMs) deliver strong performance on complex reasoning tasks, but their long chain-of-thought traces incur substantial inference latency and cost. To mitigate this overhead while preserving reasoning quality, token-level routing between a small language model (SLM) and a large language model (LLM) offers one of the best efficiency-performance trade-offs. However, existing token routers face two major practical limitations: they must be retrained for every new SLM-LLM pair, and they require both models to share a tokenizer. We present UniToR, a universal token-level router that removes both constraints. UniToR predicts per-token divergence using model-agnostic signals, enabling a single router to generalize across diverse SLM-LLM pairs without retraining. To support heterogeneous model pairs, we further introduce a token-alignment mechanism that eliminates the requirement for tokenizer compatibility, enabling seamless token-level switching across models with different tokenization schemes. Experimental results show that a single UniToR router generalizes across different model families (DeepSeek-R1, Qwen3, and Phi-4), including pairs with incompatible tokenizers. UniToR matches the accuracy of R1-32B at up to 6x lower cost and successfully operates on unseen model pairs where existing token-level routing methods fail."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="le-etal-2026-unitor">
<titleInfo>
<title>UniToR: Towards Universal Token-level Routing for Efficient Reasoning in Large Language Models</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">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>Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026</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">979-8-89176-444-6</identifier>
</relatedItem>
<abstract>Large Reasoning Models (LRMs) deliver strong performance on complex reasoning tasks, but their long chain-of-thought traces incur substantial inference latency and cost. To mitigate this overhead while preserving reasoning quality, token-level routing between a small language model (SLM) and a large language model (LLM) offers one of the best efficiency-performance trade-offs. However, existing token routers face two major practical limitations: they must be retrained for every new SLM-LLM pair, and they require both models to share a tokenizer. We present UniToR, a universal token-level router that removes both constraints. UniToR predicts per-token divergence using model-agnostic signals, enabling a single router to generalize across diverse SLM-LLM pairs without retraining. To support heterogeneous model pairs, we further introduce a token-alignment mechanism that eliminates the requirement for tokenizer compatibility, enabling seamless token-level switching across models with different tokenization schemes. Experimental results show that a single UniToR router generalizes across different model families (DeepSeek-R1, Qwen3, and Phi-4), including pairs with incompatible tokenizers. UniToR matches the accuracy of R1-32B at up to 6x lower cost and successfully operates on unseen model pairs where existing token-level routing methods fail.</abstract>
<identifier type="citekey">le-etal-2026-unitor</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.212/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>4197</start>
<end>4215</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T UniToR: Towards Universal Token-level Routing for Efficient Reasoning in Large Language Models
%A Le, Nguyen-Khang
%A Do, Truong Dinh
%A Nguyen, Le-Minh
%Y Zhang, Yue
%Y Xu, Ruifeng
%Y Purwarianti, Ayu
%Y Rong, Tong
%Y Suzuki, Jun
%S Findings of the Association for Computational Linguistics: IJCNLP-AACL 2026
%D 2026
%8 November
%I Association for Computational Linguistics
%C Hengqin, China
%@ 979-8-89176-444-6
%F le-etal-2026-unitor
%X Large Reasoning Models (LRMs) deliver strong performance on complex reasoning tasks, but their long chain-of-thought traces incur substantial inference latency and cost. To mitigate this overhead while preserving reasoning quality, token-level routing between a small language model (SLM) and a large language model (LLM) offers one of the best efficiency-performance trade-offs. However, existing token routers face two major practical limitations: they must be retrained for every new SLM-LLM pair, and they require both models to share a tokenizer. We present UniToR, a universal token-level router that removes both constraints. UniToR predicts per-token divergence using model-agnostic signals, enabling a single router to generalize across diverse SLM-LLM pairs without retraining. To support heterogeneous model pairs, we further introduce a token-alignment mechanism that eliminates the requirement for tokenizer compatibility, enabling seamless token-level switching across models with different tokenization schemes. Experimental results show that a single UniToR router generalizes across different model families (DeepSeek-R1, Qwen3, and Phi-4), including pairs with incompatible tokenizers. UniToR matches the accuracy of R1-32B at up to 6x lower cost and successfully operates on unseen model pairs where existing token-level routing methods fail.
%U https://aclanthology.org/2026.findings-aacl.212/
%P 4197-4215
Markdown (Informal)
[UniToR: Towards Universal Token-level Routing for Efficient Reasoning in Large Language Models](https://aclanthology.org/2026.findings-aacl.212/) (Le et al., Findings 2026)
ACL