@inproceedings{min-etal-2026-dalsa,
title = "{DALSA}: Difficulty-Aware Long-to-Short {LLM} Reasoning with Model Merging",
author = "Min, Dongwook and
Kang, Jeonhwi and
Jeon, Donghyeon and
Kang, Inho and
Na, Seung-Hoon",
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.272/",
pages = "6023--6040",
ISBN = "979-8-89176-439-2",
abstract = "Recent advances in large language models(LLMs) have given rise to large reasoning models that achieve strong performance through explicit multi-step reasoning. However, models often generate long reasoning chains even for simple inputs. We present Difficulty-Aware Long-to-Short LLM Reasoning with Model Merging (DALSA), an inference-time framework that adaptively controls reasoning depth based on input difficulty. DALSA integrates (i) representation-based difficulty estimation, which predicts a soft failure-rate score from the prompt hidden state, and (ii) difficulty-aware model merging, which uses a quantile remapped score to modulate the interpolation coefficient between short- and long-reasoning models. DALSA trains a lightweight probe while keeping both LLMs frozen. On mathematical benchmarks with the evaluated Qwen pairs, the accuracy{--}length trade-off depends on scale: at 7B, DALSA preserves static ACM-TA average accuracy (66.60{\%} vs. 66.61{\%}) while reducing average response length by 22.35{\%}; at 1.5B, it improves average accuracy from 56.75{\%} to 58.24{\%} with a 2.27{\%} length in- crease. Additional merging baselines, routing-threshold and merge-range sweeps, and component ablations show sensitivity to the operating configuration. Domain-matched commonsense and Llama-family diagnostics do not establish broader efficiency gains, limiting the primary empirical claim to the evaluated Qwen mathematical setting."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="min-etal-2026-dalsa">
<titleInfo>
<title>DALSA: Difficulty-Aware Long-to-Short LLM Reasoning with Model Merging</title>
</titleInfo>
<name type="personal">
<namePart type="given">Dongwook</namePart>
<namePart type="family">Min</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jeonhwi</namePart>
<namePart type="family">Kang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Donghyeon</namePart>
<namePart type="family">Jeon</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Inho</namePart>
<namePart type="family">Kang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Seung-Hoon</namePart>
<namePart type="family">Na</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>Recent advances in large language models(LLMs) have given rise to large reasoning models that achieve strong performance through explicit multi-step reasoning. However, models often generate long reasoning chains even for simple inputs. We present Difficulty-Aware Long-to-Short LLM Reasoning with Model Merging (DALSA), an inference-time framework that adaptively controls reasoning depth based on input difficulty. DALSA integrates (i) representation-based difficulty estimation, which predicts a soft failure-rate score from the prompt hidden state, and (ii) difficulty-aware model merging, which uses a quantile remapped score to modulate the interpolation coefficient between short- and long-reasoning models. DALSA trains a lightweight probe while keeping both LLMs frozen. On mathematical benchmarks with the evaluated Qwen pairs, the accuracy–length trade-off depends on scale: at 7B, DALSA preserves static ACM-TA average accuracy (66.60% vs. 66.61%) while reducing average response length by 22.35%; at 1.5B, it improves average accuracy from 56.75% to 58.24% with a 2.27% length in- crease. Additional merging baselines, routing-threshold and merge-range sweeps, and component ablations show sensitivity to the operating configuration. Domain-matched commonsense and Llama-family diagnostics do not establish broader efficiency gains, limiting the primary empirical claim to the evaluated Qwen mathematical setting.</abstract>
<identifier type="citekey">min-etal-2026-dalsa</identifier>
<location>
<url>https://aclanthology.org/2026.aacl-long.272/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>6023</start>
<end>6040</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T DALSA: Difficulty-Aware Long-to-Short LLM Reasoning with Model Merging
%A Min, Dongwook
%A Kang, Jeonhwi
%A Jeon, Donghyeon
%A Kang, Inho
%A Na, Seung-Hoon
%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 min-etal-2026-dalsa
%X Recent advances in large language models(LLMs) have given rise to large reasoning models that achieve strong performance through explicit multi-step reasoning. However, models often generate long reasoning chains even for simple inputs. We present Difficulty-Aware Long-to-Short LLM Reasoning with Model Merging (DALSA), an inference-time framework that adaptively controls reasoning depth based on input difficulty. DALSA integrates (i) representation-based difficulty estimation, which predicts a soft failure-rate score from the prompt hidden state, and (ii) difficulty-aware model merging, which uses a quantile remapped score to modulate the interpolation coefficient between short- and long-reasoning models. DALSA trains a lightweight probe while keeping both LLMs frozen. On mathematical benchmarks with the evaluated Qwen pairs, the accuracy–length trade-off depends on scale: at 7B, DALSA preserves static ACM-TA average accuracy (66.60% vs. 66.61%) while reducing average response length by 22.35%; at 1.5B, it improves average accuracy from 56.75% to 58.24% with a 2.27% length in- crease. Additional merging baselines, routing-threshold and merge-range sweeps, and component ablations show sensitivity to the operating configuration. Domain-matched commonsense and Llama-family diagnostics do not establish broader efficiency gains, limiting the primary empirical claim to the evaluated Qwen mathematical setting.
%U https://aclanthology.org/2026.aacl-long.272/
%P 6023-6040
Markdown (Informal)
[DALSA: Difficulty-Aware Long-to-Short LLM Reasoning with Model Merging](https://aclanthology.org/2026.aacl-long.272/) (Min et al., AACL-IJCNLP 2026)
ACL
- Dongwook Min, Jeonhwi Kang, Donghyeon Jeon, Inho Kang, and Seung-Hoon Na. 2026. DALSA: Difficulty-Aware Long-to-Short LLM Reasoning with Model Merging. 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 6023–6040, Hengqin, China. Association for Computational Linguistics.