@inproceedings{chen-etal-2026-dependency,
title = "Dependency-Aware Trajectory Refinement for Efficient Multi-Turn Agent Fine-Tuning",
author = "Chen, Zhuo and
Zhang, Zhen and
Wang, Xinyu and
Tu, Kewei",
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.65/",
pages = "1287--1302",
ISBN = "979-8-89176-444-6",
abstract = "Multi-turn agent trajectories often contain redundant rounds (failed tool calls, parallel sub-queries, verification-only steps) that inflate both training and inference cost. We propose viewing each trajectory as a round-level dependency DAG that exposes which rounds are globally load-bearing for the final answer, and fine-tune agents on trajectories refined through this DAG. Operating on the DAG lets us apply deterministic and interpretable edits, pruning rounds whose outputs are never consumed and merging interchangeable sibling rounds, with optional LLM-based rephrasing. Models trained on these refined trajectories consistently outperform those trained on the original trajectories at lower inference cost. Specifically, across four multi-modal QA benchmarks, our refinements improve downstream accuracy by up to ,pp over vanilla SFT (and ,pp over an LLM-deletion baseline) while reducing per-sample inference messages by up to and inference tokens by up to , translating to substantial savings in compute and serving cost. Code is released."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="chen-etal-2026-dependency">
<titleInfo>
<title>Dependency-Aware Trajectory Refinement for Efficient Multi-Turn Agent Fine-Tuning</title>
</titleInfo>
<name type="personal">
<namePart type="given">Zhuo</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zhen</namePart>
<namePart type="family">Zhang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xinyu</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Kewei</namePart>
<namePart type="family">Tu</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>Multi-turn agent trajectories often contain redundant rounds (failed tool calls, parallel sub-queries, verification-only steps) that inflate both training and inference cost. We propose viewing each trajectory as a round-level dependency DAG that exposes which rounds are globally load-bearing for the final answer, and fine-tune agents on trajectories refined through this DAG. Operating on the DAG lets us apply deterministic and interpretable edits, pruning rounds whose outputs are never consumed and merging interchangeable sibling rounds, with optional LLM-based rephrasing. Models trained on these refined trajectories consistently outperform those trained on the original trajectories at lower inference cost. Specifically, across four multi-modal QA benchmarks, our refinements improve downstream accuracy by up to ,pp over vanilla SFT (and ,pp over an LLM-deletion baseline) while reducing per-sample inference messages by up to and inference tokens by up to , translating to substantial savings in compute and serving cost. Code is released.</abstract>
<identifier type="citekey">chen-etal-2026-dependency</identifier>
<location>
<url>https://aclanthology.org/2026.findings-aacl.65/</url>
</location>
<part>
<date>2026-11</date>
<extent unit="page">
<start>1287</start>
<end>1302</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Dependency-Aware Trajectory Refinement for Efficient Multi-Turn Agent Fine-Tuning
%A Chen, Zhuo
%A Zhang, Zhen
%A Wang, Xinyu
%A Tu, Kewei
%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 chen-etal-2026-dependency
%X Multi-turn agent trajectories often contain redundant rounds (failed tool calls, parallel sub-queries, verification-only steps) that inflate both training and inference cost. We propose viewing each trajectory as a round-level dependency DAG that exposes which rounds are globally load-bearing for the final answer, and fine-tune agents on trajectories refined through this DAG. Operating on the DAG lets us apply deterministic and interpretable edits, pruning rounds whose outputs are never consumed and merging interchangeable sibling rounds, with optional LLM-based rephrasing. Models trained on these refined trajectories consistently outperform those trained on the original trajectories at lower inference cost. Specifically, across four multi-modal QA benchmarks, our refinements improve downstream accuracy by up to ,pp over vanilla SFT (and ,pp over an LLM-deletion baseline) while reducing per-sample inference messages by up to and inference tokens by up to , translating to substantial savings in compute and serving cost. Code is released.
%U https://aclanthology.org/2026.findings-aacl.65/
%P 1287-1302
Markdown (Informal)
[Dependency-Aware Trajectory Refinement for Efficient Multi-Turn Agent Fine-Tuning](https://aclanthology.org/2026.findings-aacl.65/) (Chen et al., Findings 2026)
ACL