@inproceedings{gao-etal-2026-agentcollab,
title = "{A}gent{C}ollab: A Self-Evaluation-Driven Collaboration Paradigm for Efficient {LLM} Agents",
author = "Gao, Wenbo and
Liu, Renxi and
Wang, Xian and
Fang, Guo and
Yang, Shuai and
xi, Chen and
Zhen, Hui-Ling and
Chen, Hanting and
Lin, Weizhe and
Li, Xiaosong and
Wang, Yaoyuan",
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.2400/",
pages = "54555--54570",
ISBN = "979-8-89176-459-0",
abstract = "Autonomous agents powered by large language models (LLMs) perform complex tasks through long-horizon reasoning and tool interaction, where a fundamental trade-off arises between execution efficiency and reasoning robustness. Models at different capability{--}cost levels offer complementary advantages: lower-cost models enable fast execution but may struggle on difficult reasoning segments, while stronger models provide more robust reasoning at higher computational cost. We present \textit{AgentCollab}, a self-driven collaborative inference framework that dynamically coordinates models with different reasoning capacities during agent execution. Instead of relying on external routing modules, the framework uses the agent{'}s own self-reflection signal to determine whether the current reasoning trajectory is making meaningful progress, and escalates control to a stronger reasoning tier only when necessary. To further stabilize long-horizon execution, we introduce a difficulty-aware cumulative escalation strategy that allocates additional reasoning budget based on recent failure signals. In our experiments, we instantiate this framework using a two-level small{--}large model setting. Experiments on diverse multi-step agent benchmarks show that AgentCollab consistently improves the accuracy{--}efficiency Pareto frontier of LLM agents."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="gao-etal-2026-agentcollab">
<titleInfo>
<title>AgentCollab: A Self-Evaluation-Driven Collaboration Paradigm for Efficient LLM Agents</title>
</titleInfo>
<name type="personal">
<namePart type="given">Wenbo</namePart>
<namePart type="family">Gao</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Renxi</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xian</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guo</namePart>
<namePart type="family">Fang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Shuai</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Chen</namePart>
<namePart type="family">xi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hui-Ling</namePart>
<namePart type="family">Zhen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hanting</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weizhe</namePart>
<namePart type="family">Lin</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaosong</namePart>
<namePart type="family">Li</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yaoyuan</namePart>
<namePart type="family">Wang</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>Autonomous agents powered by large language models (LLMs) perform complex tasks through long-horizon reasoning and tool interaction, where a fundamental trade-off arises between execution efficiency and reasoning robustness. Models at different capability–cost levels offer complementary advantages: lower-cost models enable fast execution but may struggle on difficult reasoning segments, while stronger models provide more robust reasoning at higher computational cost. We present AgentCollab, a self-driven collaborative inference framework that dynamically coordinates models with different reasoning capacities during agent execution. Instead of relying on external routing modules, the framework uses the agent’s own self-reflection signal to determine whether the current reasoning trajectory is making meaningful progress, and escalates control to a stronger reasoning tier only when necessary. To further stabilize long-horizon execution, we introduce a difficulty-aware cumulative escalation strategy that allocates additional reasoning budget based on recent failure signals. In our experiments, we instantiate this framework using a two-level small–large model setting. Experiments on diverse multi-step agent benchmarks show that AgentCollab consistently improves the accuracy–efficiency Pareto frontier of LLM agents.</abstract>
<identifier type="citekey">gao-etal-2026-agentcollab</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.2400/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>54555</start>
<end>54570</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T AgentCollab: A Self-Evaluation-Driven Collaboration Paradigm for Efficient LLM Agents
%A Gao, Wenbo
%A Liu, Renxi
%A Wang, Xian
%A Fang, Guo
%A Yang, Shuai
%A xi, Chen
%A Zhen, Hui-Ling
%A Chen, Hanting
%A Lin, Weizhe
%A Li, Xiaosong
%A Wang, Yaoyuan
%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 gao-etal-2026-agentcollab
%X Autonomous agents powered by large language models (LLMs) perform complex tasks through long-horizon reasoning and tool interaction, where a fundamental trade-off arises between execution efficiency and reasoning robustness. Models at different capability–cost levels offer complementary advantages: lower-cost models enable fast execution but may struggle on difficult reasoning segments, while stronger models provide more robust reasoning at higher computational cost. We present AgentCollab, a self-driven collaborative inference framework that dynamically coordinates models with different reasoning capacities during agent execution. Instead of relying on external routing modules, the framework uses the agent’s own self-reflection signal to determine whether the current reasoning trajectory is making meaningful progress, and escalates control to a stronger reasoning tier only when necessary. To further stabilize long-horizon execution, we introduce a difficulty-aware cumulative escalation strategy that allocates additional reasoning budget based on recent failure signals. In our experiments, we instantiate this framework using a two-level small–large model setting. Experiments on diverse multi-step agent benchmarks show that AgentCollab consistently improves the accuracy–efficiency Pareto frontier of LLM agents.
%U https://aclanthology.org/2026.emnlp-main.2400/
%P 54555-54570
Markdown (Informal)
[AgentCollab: A Self-Evaluation-Driven Collaboration Paradigm for Efficient LLM Agents](https://aclanthology.org/2026.emnlp-main.2400/) (Gao et al., EMNLP 2026)
ACL
- Wenbo Gao, Renxi Liu, Xian Wang, Guo Fang, Shuai Yang, Chen xi, Hui-Ling Zhen, Hanting Chen, Weizhe Lin, Xiaosong Li, and Yaoyuan Wang. 2026. AgentCollab: A Self-Evaluation-Driven Collaboration Paradigm for Efficient LLM Agents. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 54555–54570, Budapest, Hungary. Association for Computational Linguistics.