@inproceedings{jin-etal-2026-evolution,
title = "When Evolution Meets Momentum: Orchestrating Goal-oriented and Process-oriented reasoning for {LLM} Inference Scaling",
author = "Jin, Mingyu and
Chen, Haifeng and
Chen, Yuanzhou and
Light, Jonathan and
Yu, Wenchao and
Cheng, Wei",
editor = "Lee, Hung-Yi and
Dev, Sunipa and
Li, Jessy and
Elliott, Desmond",
booktitle = "Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing",
month = oct,
year = "2026",
address = "Budapest, Hungary",
publisher = "Association for Computational Linguistics",
url = "https://aclanthology.org/2026.emnlp-main.795/",
pages = "18454--18489",
ISBN = "979-8-89176-459-0",
abstract = "Large language models (LLMs) have demonstrated strong reasoning ability when given additional compute at inference time. However, existing inference-time scaling methods are fundamentally limited by their design. On the one hand, \textit{goal-oriented} approaches, such as Line or Tree Search, refine candidate solutions using previous feedback but are vulnerable to sequential dependence, often collapsing into suboptimal reasoning trajectories. On the other hand, \textit{process-oriented} approaches such as Best-of-N sampling encourage diversity through random exploration but lack feedback mechanisms, leading to inefficient computation allocation and unguided search. In this work, we propose \textbf{\textit{EvoMo}}, a novel inference-time scaling approach that unifies both paradigms by embedding a globally evolving strategy pool into MCTS, where each node expansion selects reasoning strategies under an -soft policy. To further avoid stagnation in familiar strategies, we introduce a \textit{momentum-based optimization} mechanism that monitors similarity among generated solutions and encourages the exploration of underutilized strategies. We also present a theoretical analysis of the effectiveness of the proposed momentum mechanism by modeling it as a self-evolving pool of actions. Across benchmarks, \textbf{\textit{EvoMo}} reveals significant performance gains over SOTA inference scaling methods. We observe competitive performance, like in APPS and CodeContests, with up to a 4.5{\%} and 3.63{\%} relative gain in Pass@K compared to the leading baseline within 40 search-budget, highlighting the potential of hybrid evolutionary search as a general paradigm for test-time scaling."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="jin-etal-2026-evolution">
<titleInfo>
<title>When Evolution Meets Momentum: Orchestrating Goal-oriented and Process-oriented reasoning for LLM Inference Scaling</title>
</titleInfo>
<name type="personal">
<namePart type="given">Mingyu</namePart>
<namePart type="family">Jin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haifeng</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuanzhou</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jonathan</namePart>
<namePart type="family">Light</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wenchao</namePart>
<namePart type="family">Yu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Wei</namePart>
<namePart type="family">Cheng</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>Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing</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">979-8-89176-459-0</identifier>
</relatedItem>
<abstract>Large language models (LLMs) have demonstrated strong reasoning ability when given additional compute at inference time. However, existing inference-time scaling methods are fundamentally limited by their design. On the one hand, goal-oriented approaches, such as Line or Tree Search, refine candidate solutions using previous feedback but are vulnerable to sequential dependence, often collapsing into suboptimal reasoning trajectories. On the other hand, process-oriented approaches such as Best-of-N sampling encourage diversity through random exploration but lack feedback mechanisms, leading to inefficient computation allocation and unguided search. In this work, we propose EvoMo, a novel inference-time scaling approach that unifies both paradigms by embedding a globally evolving strategy pool into MCTS, where each node expansion selects reasoning strategies under an -soft policy. To further avoid stagnation in familiar strategies, we introduce a momentum-based optimization mechanism that monitors similarity among generated solutions and encourages the exploration of underutilized strategies. We also present a theoretical analysis of the effectiveness of the proposed momentum mechanism by modeling it as a self-evolving pool of actions. Across benchmarks, EvoMo reveals significant performance gains over SOTA inference scaling methods. We observe competitive performance, like in APPS and CodeContests, with up to a 4.5% and 3.63% relative gain in Pass@K compared to the leading baseline within 40 search-budget, highlighting the potential of hybrid evolutionary search as a general paradigm for test-time scaling.</abstract>
<identifier type="citekey">jin-etal-2026-evolution</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.795/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>18454</start>
<end>18489</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T When Evolution Meets Momentum: Orchestrating Goal-oriented and Process-oriented reasoning for LLM Inference Scaling
%A Jin, Mingyu
%A Chen, Haifeng
%A Chen, Yuanzhou
%A Light, Jonathan
%A Yu, Wenchao
%A Cheng, Wei
%Y Lee, Hung-Yi
%Y Dev, Sunipa
%Y Li, Jessy
%Y Elliott, Desmond
%S Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing
%D 2026
%8 October
%I Association for Computational Linguistics
%C Budapest, Hungary
%@ 979-8-89176-459-0
%F jin-etal-2026-evolution
%X Large language models (LLMs) have demonstrated strong reasoning ability when given additional compute at inference time. However, existing inference-time scaling methods are fundamentally limited by their design. On the one hand, goal-oriented approaches, such as Line or Tree Search, refine candidate solutions using previous feedback but are vulnerable to sequential dependence, often collapsing into suboptimal reasoning trajectories. On the other hand, process-oriented approaches such as Best-of-N sampling encourage diversity through random exploration but lack feedback mechanisms, leading to inefficient computation allocation and unguided search. In this work, we propose EvoMo, a novel inference-time scaling approach that unifies both paradigms by embedding a globally evolving strategy pool into MCTS, where each node expansion selects reasoning strategies under an -soft policy. To further avoid stagnation in familiar strategies, we introduce a momentum-based optimization mechanism that monitors similarity among generated solutions and encourages the exploration of underutilized strategies. We also present a theoretical analysis of the effectiveness of the proposed momentum mechanism by modeling it as a self-evolving pool of actions. Across benchmarks, EvoMo reveals significant performance gains over SOTA inference scaling methods. We observe competitive performance, like in APPS and CodeContests, with up to a 4.5% and 3.63% relative gain in Pass@K compared to the leading baseline within 40 search-budget, highlighting the potential of hybrid evolutionary search as a general paradigm for test-time scaling.
%U https://aclanthology.org/2026.emnlp-main.795/
%P 18454-18489
Markdown (Informal)
[When Evolution Meets Momentum: Orchestrating Goal-oriented and Process-oriented reasoning for LLM Inference Scaling](https://aclanthology.org/2026.emnlp-main.795/) (Jin et al., EMNLP 2026)
ACL