AIConfPaper
← Search

Dongwon Lee

28 accepted papers

2026

Position: Breaking the Dual Curse of Multilingual AI Requires Socio-Technical Guardrails, Not Post-Hoc Alignment

Jason Lucas, Pureheart Ogheneogaga Irikefe, Adaku Uchendu, Umniya Najaer, Cornelius Adejoro, Patrice Sterling +1
ICML 2026poster

Large language models are deployed globally as universal systems, yet their safety mechanisms remain English-optimized. This creates a Dual Curse for speakers of low-resource languages: a Harmfulness Curse where harmful content generation rises from 1\% in English to 35\% in languages like Hausa, Ig…

Cited by 0Source ↗Scholar ↗
2025

Beyond Checkmate: Exploring the Creative Choke Points for AI Generated Texts

Nafis Irtiza Tripto, Saranya Venkatraman, Mahjabin Nahar, Dongwon Lee
EMNLP 2025

The rapid advancement of Large Language Models (LLMs) has revolutionized text generation but also raised concerns about potential misuse, making detecting LLM-generated text (AI text) increasingly essential. While prior work has focused on identifying AI text and effectively checkmating it, our stud

Cited by 0Source ↗PDF ↗code ↗Scholar ↗
2025

CORDIAL: Can Multimodal Large Language Models Effectively Understand Coherence Relationships?

Aashish Anantha Ramakrishnan, Aadarsh Anantha Ramakrishnan, Dongwon Lee
ACL 2025long

Multimodal Large Language Models (MLLMs) are renowned for their superior instruction-following and reasoning capabilities across diverse problem domains. However, existing benchmarks primarily focus on assessing factual and logical correctness in downstream tasks, with limited emphasis on evaluating…

Cited by 0Source ↗PDF ↗code ↗Scholar ↗
2025

CORRECT: Context- and Reference-Augmented Reasoning and Prompting for Fact-Checking

Delvin Ce Zhang, Dongwon Lee
NAACL 2025long

Fact-checking the truthfulness of claims usually requires reasoning over multiple evidence sentences. Oftentimes, evidence sentences may not be always self-contained, and may require additional contexts and references from elsewhere to understand coreferential expressions, acronyms, and the scope of…

Cited by 0Source ↗PDF ↗code ↗Scholar ↗
2025

CollabStory: Multi-LLM Collaborative Story Generation and Authorship Analysis

Saranya Venkatraman, Nafis Irtiza Tripto, Dongwon Lee
NAACL 2025findings

The rise of unifying frameworks that enable seamless interoperability of Large Language Models (LLMs) has made LLM-LLM collaboration for open-ended tasks a possibility. Despite this, there have not been efforts to explore such collaborative writing. We take the next step beyond human-LLM collaborati…

Cited by 15Source ↗PDF ↗code ↗Scholar ↗
2025

Divide-Verify-Refine: Can LLMs Self-align with Complex Instructions?

Xianren Zhang, Xianfeng Tang, Hui Liu, Zongyu Wu, Qi He, Dongwon Lee +1
ACL 2025finding

Recent studies show LLMs struggle with complex instructions involving multiple constraints (e.g., length, format, sentiment). Existing research enhances open-source LLMs using closed-source guidance (e.g., GPT-4), but this heavily relies on generated data quality. An alternative is leveraging LLMs’…

Cited by 0Source ↗PDF ↗Scholar ↗
2025

From Intentions to Techniques: A Comprehensive Taxonomy and Challenges in Text Watermarking for Large Language Models

Harsh Nishant Lalai, Aashish Anantha Ramakrishnan, Raj Sanjay Shah, Dongwon Lee
NAACL 2025findings

With the rapid growth of Large Language Models (LLMs), safeguarding textual content against unauthorized use is crucial. Watermarking offers a vital solution, protecting both - LLM-generated and plain text sources. This paper presents a unified overview of different perspectives behind designing wat…

Cited by 2Source ↗PDF ↗Scholar ↗
2025

LaMP-Cap: Personalized Figure Caption Generation With Multimodal Figure Profiles

Ho Yin Sam Ng, Edward Hsu, Aashish Anantha Ramakrishnan, Branislav Kveton, Nedim Lipka, Franck Dernoncourt +5
EMNLP 2025

Figure captions are crucial for helping readers understand and remember a figure’s key message. Many models have been developed to generate these captions, helping authors compose better quality captions more easily. Yet, authors almost always need to revise generic AI-generated captions to match th

Cited by 0Source ↗PDF ↗code ↗Scholar ↗
2025

PlagBench: Exploring the Duality of Large Language Models in Plagiarism Generation and Detection

Jooyoung Lee, Toshini Agrawal, Adaku Uchendu, Thai Le, Jinghui Chen, Dongwon Lee
NAACL 2025long

Recent studies have raised concerns about the potential threats large language models (LLMs) pose to academic integrity and copyright protection. Yet, their investigation is predominantly focused on literal copies of original texts. Also, how LLMs can facilitate the detection of LLM-generated plagia…

Cited by 5Source ↗PDF ↗code ↗Scholar ↗
2025

SUA: Stealthy Multimodal Large Language Model Unlearning Attack

Xianren Zhang, Hui Liu, Delvin Ce Zhang, Xianfeng Tang, Qi He, Dongwon Lee +1
EMNLP 2025

Multimodal Large Language Models (MLLMs) trained on massive data may memorize sensitive personal information and photos, posing serious privacy risks. To mitigate this, MLLM unlearning methods are proposed, which fine-tune MLLMs to reduce the “forget” sensitive information. However, it remains uncle

Cited by 0Source ↗PDF ↗Scholar ↗
2025

Unmasking Fake Careers: Detecting Machine-Generated Career Trajectories via Multi-layer Heterogeneous Graphs

Michiharu Yamashita, Thanh Tran, Delvin Ce Zhang, Dongwon Lee
EMNLP 2025

The rapid advancement of Large Language Models (LLMs) has enabled the generation of highly realistic synthetic data. We identify a new vulnerability, LLMs generating convincing career trajectories in fake resumes and explore effective detection methods. To address this challenge, we construct a data

Cited by 0Source ↗PDF ↗code ↗Scholar ↗
2024

A Ship of Theseus: Curious Cases of Paraphrasing in LLM-Generated Texts

Nafis Irtiza Tripto, Saranya Venkatraman, Dominik Macko, Robert Moro, Ivan Srba, Adaku Uchendu +2
ACL 2024long

In the realm of text manipulation and linguistic transformation, the question of authorship has been a subject of fascination and philosophical inquiry. Much like the Ship of Theseus paradox, which ponders whether a ship remains the same when each of its original planks is replaced, our research del…

Cited by 4Source ↗PDF ↗code ↗Scholar ↗
2024

ALISON: Fast and Effective Stylometric Authorship Obfuscation

Eric Xing, Saranya Venkatraman, Thai Le, Dongwon Lee
AAAI 2024technical

Authorship Attribution (AA) and Authorship Obfuscation (AO) are two competing tasks of increasing importance in privacy research. Modern AA leverages an author's consistent writing style to match a text to its author using an AA classifier. AO is the corresponding adversarial task, aiming to modify…

Cited by 4Source ↗PDF ↗code ↗Scholar ↗
2024

Authorship Obfuscation in Multilingual Machine-Generated Text Detection

Dominik Macko, Robert Moro, Adaku Uchendu, Ivan Srba, Jason S Lucas, Michiharu Yamashita +4
EMNLP 2024finding

High-quality text generation capability of latest Large Language Models (LLMs) causes concerns about their misuse (e.g., in massive generation/spread of disinformation). Machine-generated text (MGT) detection is important to cope with such threats. However, it is susceptible to authorship obfuscatio…

Cited by 21Source ↗PDF ↗code ↗Scholar ↗
2024

Comprehensive Attribution: Inherently Explainable Vision Model with Feature Detector

Xianren Zhang, Dongwon Lee, Suhang Wang*
ECCV 2024poster

"As deep vision models’ popularity rapidly increases, there is a growing emphasis on explanations for model predictions. The inherently explainable attribution method aims to enhance the understanding of model behavior by identifying the important regions in images that significantly contribute to p…

Cited by 2Source ↗PDF ↗code ↗Scholar ↗
2024

GPT-who: An Information Density-based Machine-Generated Text Detector

Saranya Venkatraman, Adaku Uchendu, Dongwon Lee
NAACL 2024findings

The Uniform Information Density (UID) principle posits that humans prefer to spread information evenly during language production. We examine if this UID principle can help capture differences between Large Language Models (LLMs)-generated and human-generated texts. We propose GPT-who, the first psy…

Cited by 36Source ↗PDF ↗code ↗Scholar ↗
2023

Fighting Fire with Fire: The Dual Role of LLMs in Crafting and Detecting Elusive Disinformation

Jason S Lucas, Adaku Uchendu, Michiharu Yamashita, Jooyoung Lee, Shaurya Rohatgi, Dongwon Lee
EMNLP 2023long main

Recent ubiquity and disruptive impacts of large language models (LLMs) have raised concerns about their potential to be misused (*.i.e, generating large-scale harmful and misleading content*). To combat this emerging risk of LLMs, we propose a novel "***Fighting Fire with Fire***" (F3) strategy that…

Cited by 0Source ↗code ↗Scholar ↗
2023

HANSEN: Human and AI Spoken Text Benchmark for Authorship Analysis

Nafis Irtiza Tripto, Adaku Uchendu, Thai Le, Mattia Setzu, Fosca Giannotti, Dongwon Lee
EMNLP 2023long findings

$\textit{Authorship Analysis}$, also known as stylometry, has been an essential aspect of Natural Language Processing (NLP) for a long time. Likewise, the recent advancement of Large Language Models (LLMs) has made authorship analysis increasingly crucial for distinguishing between human-written and…

Cited by 0Source ↗Scholar ↗
2023

MULTITuDE: Large-Scale Multilingual Machine-Generated Text Detection Benchmark

Dominik Macko, Robert Moro, Adaku Uchendu, Jason S Lucas, Michiharu Yamashita, Matúš Pikuliak +5
EMNLP 2023long main

There is a lack of research into capabilities of recent LLMs to generate convincing text in languages other than English and into performance of detectors of machine-generated text in multilingual settings. This is also reflected in the available benchmarks which lack authentic texts in languages ot…

Cited by 0Source ↗code ↗Scholar ↗
2023

UPTON: Preventing Authorship Leakage from Public Text Release via Data Poisoning

Ziyao Wang, Thai Le, Dongwon Lee
EMNLP 2023long findings

Consider a scenario where an author (e.g., activist, whistle-blower) with many public writings wishes to write “anonymously" when attackers may have already built an authorship attribution (AA) model based off of public writings including those of the author. To enable her wish, we ask a question “c…

Cited by 0Source ↗Scholar ↗
2022

Forecasting the Number of Tenants At-Risk of Formal Eviction: A Machine Learning Approach to Inform Public Policy

Maryam Tabar, Wooyong Jung, Amulya Yadav, Owen Wilson Chavez, Ashley Flores, Dongwon Lee
IJCAI 2022poster

Eviction of tenants has reached a crisis level in the U.S. and its consequences pose significant challenges to society. To tackle this eviction crisis, policymakers have been allocating financial resources but a more efficient resource allocation would need an accurate forecast of the number of tena…

Cited by 8Source ↗PDF ↗code ↗Scholar ↗
2022

Perturbations in the Wild: Leveraging Human-Written Text Perturbations for Realistic Adversarial Attack and Defense

Thai Le, Jooyoung Lee, Kevin Yen, Yifan Hu, Dongwon Lee
ACL 2022findings

We proposes a novel algorithm, ANTHRO, that inductively extracts over 600K human-written text perturbations in the wild and leverages them for realistic adversarial attack. Unlike existing character-based attacks which often deductively hypothesize a set of manipulation strategies, our work is groun…

Cited by 38Source ↗PDF ↗code ↗Scholar ↗
2022

SHIELD: Defending Textual Neural Networks against Multiple Black-Box Adversarial Attacks with Stochastic Multi-Expert Patcher

Thai Le, Noseong Park, Dongwon Lee
ACL 2022long

Even though several methods have proposed to defend textual neural network (NN) models against black-box adversarial attacks, they often defend against a specific text perturbation strategy and/or require re-training the models from scratch. This leads to a lack of generalization in practice and red…

Cited by 24Source ↗PDF ↗code ↗Scholar ↗
2021

A Sweet Rabbit Hole by DARCY: Using Honeypots to Detect Universal Trigger’s Adversarial Attacks

Thai Le, Noseong Park, Dongwon Lee
ACL 2021long

The Universal Trigger (UniTrigger) is a recently-proposed powerful adversarial textual attack method. Utilizing a learning-based mechanism, UniTrigger generates a fixed phrase that, when added to any benign inputs, can drop the prediction accuracy of a textual neural network (NN) model to near zero…

Cited by 28Source ↗PDF ↗code ↗Scholar ↗
2021

MelBERT: Metaphor Detection via Contextualized Late Interaction using Metaphorical Identification Theories

Minjin Choi, Sunkyung Lee, Eunseong Choi, Heesoo Park, Junhyuk Lee, Dongwon Lee +1
NAACL 2021long

Automated metaphor detection is a challenging task to identify the metaphorical expression of words in a sentence. To tackle this problem, we adopt pre-trained contextualized models, e.g., BERT and RoBERTa. To this end, we propose a novel metaphor detection model, namely metaphor-aware late interact…

Cited by 110Source ↗PDF ↗code ↗Scholar ↗
2021

TURINGBENCH: A Benchmark Environment for Turing Test in the Age of Neural Text Generation

Adaku Uchendu, Zeyu Ma, Thai Le, Rui Zhang, Dongwon Lee
EMNLP 2021finding

Recent progress in generative language models has enabled machines to generate astonishingly realistic texts. While there are many legitimate applications of such models, there is also a rising need to distinguish machine-generated texts from human-written ones (e.g., fake news detection). However,…

Cited by 125Source ↗PDF ↗code ↗Scholar ↗