@inproceedings{li-etal-2026-best,
title = "The Best of Both Worlds: Combining Parallel and Sequential Inference Scaling via Aggregation Fine-Tuning",
author = "Li, Yafu and
Wang, Zhilin and
Fu, Tingchen and
Cui, Ganqu and
Yang, Sen and
Cheng, Yu",
editor = "Liakata, Maria and
Moreira, Viviane P. and
Zhang, Jiajun and
Jurgens, David",
booktitle = "Findings of the {A}ssociation for {C}omputational {L}inguistics: {ACL} 2026",
month = jul,
year = "2026",
address = "San Diego, California, United States",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.findings-acl.1568/",
pages = "31369--31389",
ISBN = "979-8-89176-395-1",
abstract = "Scaling data and model size has been proven effective for boosting the performance of large language models. In addition to training-time scaling, recent studies have revealed that increasing test-time computational resources can further improve performance. In this work, we introduce Aggregation Fine-Tuning (AFT), a supervised fine-tuning paradigm where the model learns to synthesize multiple draft responses, referred to as proposals, into a single, refined answer, termed aggregation. At inference time, we apply a propose-and-aggregate strategy that iteratively generates and aggregates proposals, effectively scaling inference-time computation without relying on external guidance such as a reward model. Empirical results across benchmark datasets demonstrate that AFT-trained models achieve substantial gains with test-time scaling, outperforming best-of-N baselines while eliminating the need for external reward signals. Notably, an AFT model, fine-tuned from Llama3.1-8B-Base with only 64k data, achieves a 41.3{\%} LC win rate on AlpacaEval 2, surpassing significantly larger LLMs such as Llama3.1-405B-Instruct and GPT-4. By combining sequential refinement and parallel sampling, the propose-and-aggregate framework scales inference-time computation in a flexible manner."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="li-etal-2026-best">
<titleInfo>
<title>The Best of Both Worlds: Combining Parallel and Sequential Inference Scaling via Aggregation Fine-Tuning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yafu</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhilin</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tingchen</namePart>
<namePart type="family">Fu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Ganqu</namePart>
<namePart type="family">Cui</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Sen</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yu</namePart>
<namePart type="family">Cheng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<originInfo>
<dateIssued>2026-07</dateIssued>
</originInfo>
<typeOfResource>text</typeOfResource>
<relatedItem type="host">
<titleInfo>
<title>Findings of the Association for Computational Linguistics: ACL 2026</title>
</titleInfo>
<name type="personal">
<namePart type="given">Maria</namePart>
<namePart type="family">Liakata</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Viviane</namePart>
<namePart type="given">P</namePart>
<namePart type="family">Moreira</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiajun</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">David</namePart>
<namePart type="family">Jurgens</namePart>
<role>
<roleTerm authority="marcrelator" type="text">editor</roleTerm>
</role>
</name>
<originInfo>
<publisher>Association for Computational Linguistics</publisher>
<place>
<placeTerm type="text">San Diego, California, United States</placeTerm>
</place>
</originInfo>
<genre authority="marcgt">conference publication</genre>
<identifier type="isbn">979-8-89176-395-1</identifier>
</relatedItem>
<abstract>Scaling data and model size has been proven effective for boosting the performance of large language models. In addition to training-time scaling, recent studies have revealed that increasing test-time computational resources can further improve performance. In this work, we introduce Aggregation Fine-Tuning (AFT), a supervised fine-tuning paradigm where the model learns to synthesize multiple draft responses, referred to as proposals, into a single, refined answer, termed aggregation. At inference time, we apply a propose-and-aggregate strategy that iteratively generates and aggregates proposals, effectively scaling inference-time computation without relying on external guidance such as a reward model. Empirical results across benchmark datasets demonstrate that AFT-trained models achieve substantial gains with test-time scaling, outperforming best-of-N baselines while eliminating the need for external reward signals. Notably, an AFT model, fine-tuned from Llama3.1-8B-Base with only 64k data, achieves a 41.3% LC win rate on AlpacaEval 2, surpassing significantly larger LLMs such as Llama3.1-405B-Instruct and GPT-4. By combining sequential refinement and parallel sampling, the propose-and-aggregate framework scales inference-time computation in a flexible manner.</abstract>
<identifier type="citekey">li-etal-2026-best</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.1568/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>31369</start>
<end>31389</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T The Best of Both Worlds: Combining Parallel and Sequential Inference Scaling via Aggregation Fine-Tuning
%A Li, Yafu
%A Wang, Zhilin
%A Fu, Tingchen
%A Cui, Ganqu
%A Yang, Sen
%A Cheng, Yu
%Y Liakata, Maria
%Y Moreira, Viviane P.
%Y Zhang, Jiajun
%Y Jurgens, David
%S Findings of the Association for Computational Linguistics: ACL 2026
%D 2026
%8 July
%I Association for Computational Linguistics
%C San Diego, California, United States
%@ 979-8-89176-395-1
%F li-etal-2026-best
%X Scaling data and model size has been proven effective for boosting the performance of large language models. In addition to training-time scaling, recent studies have revealed that increasing test-time computational resources can further improve performance. In this work, we introduce Aggregation Fine-Tuning (AFT), a supervised fine-tuning paradigm where the model learns to synthesize multiple draft responses, referred to as proposals, into a single, refined answer, termed aggregation. At inference time, we apply a propose-and-aggregate strategy that iteratively generates and aggregates proposals, effectively scaling inference-time computation without relying on external guidance such as a reward model. Empirical results across benchmark datasets demonstrate that AFT-trained models achieve substantial gains with test-time scaling, outperforming best-of-N baselines while eliminating the need for external reward signals. Notably, an AFT model, fine-tuned from Llama3.1-8B-Base with only 64k data, achieves a 41.3% LC win rate on AlpacaEval 2, surpassing significantly larger LLMs such as Llama3.1-405B-Instruct and GPT-4. By combining sequential refinement and parallel sampling, the propose-and-aggregate framework scales inference-time computation in a flexible manner.
%U https://aclanthology.org/2026.findings-acl.1568/
%P 31369-31389
Markdown (Informal)
[The Best of Both Worlds: Combining Parallel and Sequential Inference Scaling via Aggregation Fine-Tuning](https://aclanthology.org/2026.findings-acl.1568/) (Li et al., Findings 2026)
ACL