@inproceedings{huang-etal-2026-mdhb,
title = "{MDHB}: A Multi-Dimensional Online Harm Detection Benchmark for Proactive Psychosocial Risk Forecasting",
author = "Huang, Yuting and
Huang, Xiaoyong and
Wang, Hao and
Sun, Heli and
Wu, Jiaruo and
Wang, Zejun and
Shi, Yunyun and
Yue, Guanlan and
He, Liang",
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.1210/",
pages = "27663--27679",
ISBN = "979-8-89176-459-0",
abstract = "Online harm detection is crucial for safeguarding digital ecosystems. However, current research is largely confined to single-dimensional classification paradigms or generative safety alignment under simplified and explicit adversarial prompts. To capture complex and highly covert threats in the real world, we introduce the Multi-dimensional Online Harm Detection Benchmark (MDHB). MDHB focuses on four core dimensions critically demanded by social platforms: misinformation, toxicity, sarcasm, and logical fallacies. Notably, we incorporate non-offensive implicit sarcasm (e.g., self-deprecation and veiled complaints) to propel content governance toward proactive psychosocial risk forecasting. Evaluations reveal that mainstream Large Language Models (LLMs), including GPT-5, suffer from severe cross-dimensional interference under both direct inference and ReAct paradigms. To address this, we propose the Self-Refining Sentinel (SRS), an error-driven self-evolving agent architecture. Without requiring parameter fine-tuning, SRS dynamically constructs and updates an experience repository from failure cases through an iterative self-evolution mechanism. Experimental results demonstrate that SRS significantly outperforms state-of-the-art baselines in overall performance, effectively mitigating cross-dimensional interference in multi-dimensional online threat detection."
}<?xml version="1.0" encoding="UTF-8"?>
<modsCollection xmlns="http://www.loc.gov/mods/v3">
<mods ID="huang-etal-2026-mdhb">
<titleInfo>
<title>MDHB: A Multi-Dimensional Online Harm Detection Benchmark for Proactive Psychosocial Risk Forecasting</title>
</titleInfo>
<name type="personal">
<namePart type="given">Yuting</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Xiaoyong</namePart>
<namePart type="family">Huang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Hao</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Heli</namePart>
<namePart type="family">Sun</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Jiaruo</namePart>
<namePart type="family">Wu</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Zejun</namePart>
<namePart type="family">Wang</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Yunyun</namePart>
<namePart type="family">Shi</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Guanlan</namePart>
<namePart type="family">Yue</namePart>
<role>
<roleTerm authority="marcrelator" type="text">author</roleTerm>
</role>
</name>
<name type="personal">
<namePart type="given">Liang</namePart>
<namePart type="family">He</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>Online harm detection is crucial for safeguarding digital ecosystems. However, current research is largely confined to single-dimensional classification paradigms or generative safety alignment under simplified and explicit adversarial prompts. To capture complex and highly covert threats in the real world, we introduce the Multi-dimensional Online Harm Detection Benchmark (MDHB). MDHB focuses on four core dimensions critically demanded by social platforms: misinformation, toxicity, sarcasm, and logical fallacies. Notably, we incorporate non-offensive implicit sarcasm (e.g., self-deprecation and veiled complaints) to propel content governance toward proactive psychosocial risk forecasting. Evaluations reveal that mainstream Large Language Models (LLMs), including GPT-5, suffer from severe cross-dimensional interference under both direct inference and ReAct paradigms. To address this, we propose the Self-Refining Sentinel (SRS), an error-driven self-evolving agent architecture. Without requiring parameter fine-tuning, SRS dynamically constructs and updates an experience repository from failure cases through an iterative self-evolution mechanism. Experimental results demonstrate that SRS significantly outperforms state-of-the-art baselines in overall performance, effectively mitigating cross-dimensional interference in multi-dimensional online threat detection.</abstract>
<identifier type="citekey">huang-etal-2026-mdhb</identifier>
<location>
<url>https://aclanthology.org/2026.emnlp-main.1210/</url>
</location>
<part>
<date>2026-10</date>
<extent unit="page">
<start>27663</start>
<end>27679</end>
</extent>
</part>
</mods>
</modsCollection>
%0 Conference Proceedings
%T MDHB: A Multi-Dimensional Online Harm Detection Benchmark for Proactive Psychosocial Risk Forecasting
%A Huang, Yuting
%A Huang, Xiaoyong
%A Wang, Hao
%A Sun, Heli
%A Wu, Jiaruo
%A Wang, Zejun
%A Shi, Yunyun
%A Yue, Guanlan
%A He, Liang
%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 huang-etal-2026-mdhb
%X Online harm detection is crucial for safeguarding digital ecosystems. However, current research is largely confined to single-dimensional classification paradigms or generative safety alignment under simplified and explicit adversarial prompts. To capture complex and highly covert threats in the real world, we introduce the Multi-dimensional Online Harm Detection Benchmark (MDHB). MDHB focuses on four core dimensions critically demanded by social platforms: misinformation, toxicity, sarcasm, and logical fallacies. Notably, we incorporate non-offensive implicit sarcasm (e.g., self-deprecation and veiled complaints) to propel content governance toward proactive psychosocial risk forecasting. Evaluations reveal that mainstream Large Language Models (LLMs), including GPT-5, suffer from severe cross-dimensional interference under both direct inference and ReAct paradigms. To address this, we propose the Self-Refining Sentinel (SRS), an error-driven self-evolving agent architecture. Without requiring parameter fine-tuning, SRS dynamically constructs and updates an experience repository from failure cases through an iterative self-evolution mechanism. Experimental results demonstrate that SRS significantly outperforms state-of-the-art baselines in overall performance, effectively mitigating cross-dimensional interference in multi-dimensional online threat detection.
%U https://aclanthology.org/2026.emnlp-main.1210/
%P 27663-27679
Markdown (Informal)
[MDHB: A Multi-Dimensional Online Harm Detection Benchmark for Proactive Psychosocial Risk Forecasting](https://aclanthology.org/2026.emnlp-main.1210/) (Huang et al., EMNLP 2026)
ACL
- Yuting Huang, Xiaoyong Huang, Hao Wang, Heli Sun, Jiaruo Wu, Zejun Wang, Yunyun Shi, Guanlan Yue, and Liang He. 2026. MDHB: A Multi-Dimensional Online Harm Detection Benchmark for Proactive Psychosocial Risk Forecasting. In Proceedings of the 2026 Conference on Empirical Methods in Natural Language Processing, pages 27663–27679, Budapest, Hungary. Association for Computational Linguistics.