@inproceedings{su-etal-2026-failure,
title = "Failure makes the agent stronger: Enhancing Accuracy through Structured Reflection for Reliable Tool Interactions",
author = "Su, Junhao and
Wan, Yuanliang and
Yang, Junwei and
Shi, Hengyu and
Han, Tianyang and
Qiu, Yurui and
Luo, Junfeng",
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.618/",
doi = "10.18653/v1/2026.findings-acl.618",
pages = "12712--12734",
ISBN = "979-8-89176-395-1",
abstract = "Tool-augmented large language models (LLMs) are typically trained via supervised imitation learning or coarse-grained reinforcement learning, approaches that primarily optimize one-shot tool calls. Existing practices of self-reflection largely rely on heuristic prompting or unidirectional reasoning traces: the model is encouraged to ``think more,'' rather than to treat error diagnosis and correction as a learnable capability. This makes them fragile in multi-turn interaction settings{---}once a call fails, the model tends to repeat the same mistake instead of recovering. To address this issue, we propose structured reflection, which transforms the ``from error to repair'' process into a first-class, controllable, and trainable action. The agent produces a concise yet precise reflection process: specifically, the model diagnoses the error based on evidence from the previous step and then proposes a correct and executable follow-up call. During training, we combine DAPO and GSPO{'}s objective functions and design a more principled reward mechanism tailored to tool calling, optimizing the stepwise strategy Reflect $\to$ Call $\to$ Final. To evaluate this capability, we introduce Tool-Reflection-Bench, a lightweight benchmark dataset that programmatically verifies structural validity, executability, parameter correctness, and result consistency. Tasks in the benchmark are constructed as miniature trajectories of Erroneous Call $\to$ Reflection $\to$ Corrected Call and are split into disjoint training and testing sets. Experiments on BFCL v3 and Tool-Reflection-Bench show that our method achieves significant improvements in multi-turn tool-call success rates and error recovery, while also reducing redundant calls. These results demonstrate that making reflection explicit and treating it as an optimization objective can substantially enhance the reliability of tool interaction, providing a reproducible pathway for agents to grow stronger by learning from failure. We will release all the code and datasets as open source once the paper is accepted by the community."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="su-etal-2026-failure">
<titleInfo>
<title>Failure makes the agent stronger: Enhancing Accuracy through Structured Reflection for Reliable Tool Interactions</title>
</titleInfo>
<name type="personal">
<namePart type="given">Junhao</namePart>
<namePart type="family">Su</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yuanliang</namePart>
<namePart type="family">Wan</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junwei</namePart>
<namePart type="family">Yang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hengyu</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Tianyang</namePart>
<namePart type="family">Han</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yurui</namePart>
<namePart type="family">Qiu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Junfeng</namePart>
<namePart type="family">Luo</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>Tool-augmented large language models (LLMs) are typically trained via supervised imitation learning or coarse-grained reinforcement learning, approaches that primarily optimize one-shot tool calls. Existing practices of self-reflection largely rely on heuristic prompting or unidirectional reasoning traces: the model is encouraged to “think more,” rather than to treat error diagnosis and correction as a learnable capability. This makes them fragile in multi-turn interaction settings—once a call fails, the model tends to repeat the same mistake instead of recovering. To address this issue, we propose structured reflection, which transforms the “from error to repair” process into a first-class, controllable, and trainable action. The agent produces a concise yet precise reflection process: specifically, the model diagnoses the error based on evidence from the previous step and then proposes a correct and executable follow-up call. During training, we combine DAPO and GSPO’s objective functions and design a more principled reward mechanism tailored to tool calling, optimizing the stepwise strategy Reflect Call Final. To evaluate this capability, we introduce Tool-Reflection-Bench, a lightweight benchmark dataset that programmatically verifies structural validity, executability, parameter correctness, and result consistency. Tasks in the benchmark are constructed as miniature trajectories of Erroneous Call Reflection Corrected Call and are split into disjoint training and testing sets. Experiments on BFCL v3 and Tool-Reflection-Bench show that our method achieves significant improvements in multi-turn tool-call success rates and error recovery, while also reducing redundant calls. These results demonstrate that making reflection explicit and treating it as an optimization objective can substantially enhance the reliability of tool interaction, providing a reproducible pathway for agents to grow stronger by learning from failure. We will release all the code and datasets as open source once the paper is accepted by the community.</abstract>
<identifier type="citekey">su-etal-2026-failure</identifier>
<identifier type="doi">10.18653/v1/2026.findings-acl.618</identifier>
<location>
<url>https://aclanthology.org/2026.findings-acl.618/</url>
</location>
<part>
<date>2026-07</date>
<extent unit="page">
<start>12712</start>
<end>12734</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T Failure makes the agent stronger: Enhancing Accuracy through Structured Reflection for Reliable Tool Interactions
%A Su, Junhao
%A Wan, Yuanliang
%A Yang, Junwei
%A Shi, Hengyu
%A Han, Tianyang
%A Qiu, Yurui
%A Luo, Junfeng
%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 su-etal-2026-failure
%X Tool-augmented large language models (LLMs) are typically trained via supervised imitation learning or coarse-grained reinforcement learning, approaches that primarily optimize one-shot tool calls. Existing practices of self-reflection largely rely on heuristic prompting or unidirectional reasoning traces: the model is encouraged to “think more,” rather than to treat error diagnosis and correction as a learnable capability. This makes them fragile in multi-turn interaction settings—once a call fails, the model tends to repeat the same mistake instead of recovering. To address this issue, we propose structured reflection, which transforms the “from error to repair” process into a first-class, controllable, and trainable action. The agent produces a concise yet precise reflection process: specifically, the model diagnoses the error based on evidence from the previous step and then proposes a correct and executable follow-up call. During training, we combine DAPO and GSPO’s objective functions and design a more principled reward mechanism tailored to tool calling, optimizing the stepwise strategy Reflect Call Final. To evaluate this capability, we introduce Tool-Reflection-Bench, a lightweight benchmark dataset that programmatically verifies structural validity, executability, parameter correctness, and result consistency. Tasks in the benchmark are constructed as miniature trajectories of Erroneous Call Reflection Corrected Call and are split into disjoint training and testing sets. Experiments on BFCL v3 and Tool-Reflection-Bench show that our method achieves significant improvements in multi-turn tool-call success rates and error recovery, while also reducing redundant calls. These results demonstrate that making reflection explicit and treating it as an optimization objective can substantially enhance the reliability of tool interaction, providing a reproducible pathway for agents to grow stronger by learning from failure. We will release all the code and datasets as open source once the paper is accepted by the community.
%R 10.18653/v1/2026.findings-acl.618
%U https://aclanthology.org/2026.findings-acl.618/
%U https://doi.org/10.18653/v1/2026.findings-acl.618
%P 12712-12734
Markdown (Informal)
[Failure makes the agent stronger: Enhancing Accuracy through Structured Reflection for Reliable Tool Interactions](https://aclanthology.org/2026.findings-acl.618/) (Su et al., Findings 2026)
ACL
- Junhao Su, Yuanliang Wan, Junwei Yang, Hengyu Shi, Tianyang Han, Yurui Qiu, and Junfeng Luo. 2026. Failure makes the agent stronger: Enhancing Accuracy through Structured Reflection for Reliable Tool Interactions. In Findings of the Association for Computational Linguistics: ACL 2026, pages 12712–12734, San Diego, California, United States. Association for Computational Linguistics.