@inproceedings{wang-etal-2026-captchamind,
title = "{C}aptcha{M}ind: Training {CAPTCHA} Solvers via Reinforcement Learning with Explicit Reasoning Supervision",
author = "Wang, Pengcheng and
Liu, Haoxiang and
Dai, Yang and
Zeng, Xiangxiang and
Chen, Guanhua and
Hu, Baotian and
Wang, Longyue and
Luo, Weihua",
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.1140/",
pages = "26142--26158",
ISBN = "979-8-89176-459-0",
abstract = "CAPTCHAs are widely deployed as human verification mechanisms and frequently block intelligent agents from completing end-to-end automation in real-world web environments. Existing CAPTCHA-solving methods rely on prompting closed-source models, incurring high inference costs and lacking controllability, yet training-based approaches have remained absent due to the lack of large-scale training data with process-level annotations. To enable training-based CAPTCHA solving, we introduce CaptchaBench, a large-scale benchmark of 16,000 programmatically generated samples with detailed region and process-level annotations. Systematic evaluation on CaptchaBench reveals that existing methods fail on tasks requiring fine-grained visual detail capture, with a key challenge being the inability to supervise the model{'}s reasoning process over task-relevant visual regions. Therefore, we present CaptchaMind, an RL-based solver that explicitly supervises the model{'}s reasoning process by directly rewarding correct identification of task-relevant visual regions at intermediate steps. CaptchaMind achieves state-of-the-art performance across eight tasks and generalizes effectively to real-world CAPTCHA instances, substantially outperforming all existing methods without closed-source APIs."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="wang-etal-2026-captchamind">
<titleInfo>
<title>CaptchaMind: Training CAPTCHA Solvers via Reinforcement Learning with Explicit Reasoning Supervision</title>
</titleInfo>
<name type="personal">
<namePart type="given">Pengcheng</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Haoxiang</namePart>
<namePart type="family">Liu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yang</namePart>
<namePart type="family">Dai</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiangxiang</namePart>
<namePart type="family">Zeng</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guanhua</namePart>
<namePart type="family">Chen</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Baotian</namePart>
<namePart type="family">Hu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Longyue</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Weihua</namePart>
<namePart type="family">Luo</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>CAPTCHAs are widely deployed as human verification mechanisms and frequently block intelligent agents from completing end-to-end automation in real-world web environments. Existing CAPTCHA-solving methods rely on prompting closed-source models, incurring high inference costs and lacking controllability, yet training-based approaches have remained absent due to the lack of large-scale training data with process-level annotations. To enable training-based CAPTCHA solving, we introduce CaptchaBench, a large-scale benchmark of 16,000 programmatically generated samples with detailed region and process-level annotations. Systematic evaluation on CaptchaBench reveals that existing methods fail on tasks requiring fine-grained visual detail capture, with a key challenge being the inability to supervise the model’s reasoning process over task-relevant visual regions. Therefore, we present CaptchaMind, an RL-based solver that explicitly supervises the model’s reasoning process by directly rewarding correct identification of task-relevant visual regions at intermediate steps. CaptchaMind achieves state-of-the-art performance across eight tasks and generalizes effectively to real-world CAPTCHA instances, substantially outperforming all existing methods without closed-source APIs.</abstract>
<identifier type="citekey">wang-etal-2026-captchamind</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1140/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>26142</start>
<end>26158</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T CaptchaMind: Training CAPTCHA Solvers via Reinforcement Learning with Explicit Reasoning Supervision
%A Wang, Pengcheng
%A Liu, Haoxiang
%A Dai, Yang
%A Zeng, Xiangxiang
%A Chen, Guanhua
%A Hu, Baotian
%A Wang, Longyue
%A Luo, Weihua
%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 wang-etal-2026-captchamind
%X CAPTCHAs are widely deployed as human verification mechanisms and frequently block intelligent agents from completing end-to-end automation in real-world web environments. Existing CAPTCHA-solving methods rely on prompting closed-source models, incurring high inference costs and lacking controllability, yet training-based approaches have remained absent due to the lack of large-scale training data with process-level annotations. To enable training-based CAPTCHA solving, we introduce CaptchaBench, a large-scale benchmark of 16,000 programmatically generated samples with detailed region and process-level annotations. Systematic evaluation on CaptchaBench reveals that existing methods fail on tasks requiring fine-grained visual detail capture, with a key challenge being the inability to supervise the model’s reasoning process over task-relevant visual regions. Therefore, we present CaptchaMind, an RL-based solver that explicitly supervises the model’s reasoning process by directly rewarding correct identification of task-relevant visual regions at intermediate steps. CaptchaMind achieves state-of-the-art performance across eight tasks and generalizes effectively to real-world CAPTCHA instances, substantially outperforming all existing methods without closed-source APIs.
%U https://aclanthology.org/2026.emnlp-main.1140/
%P 26142-26158
Markdown (Informal)
[CaptchaMind: Training CAPTCHA Solvers via Reinforcement Learning with Explicit Reasoning Supervision](https://aclanthology.org/2026.emnlp-main.1140/) (Wang et al., EMNLP 2026)
ACL
- Pengcheng Wang, Haoxiang Liu, Yang Dai, Xiangxiang Zeng, Guanhua Chen, Baotian Hu, Longyue Wang, and Weihua Luo. 2026. CaptchaMind: Training CAPTCHA Solvers via Reinforcement Learning with Explicit Reasoning Supervision. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 26142–26158, Budapest, Hungary. Association for Computational Linguistics.