Papers by Yejin Lee
Beyond Task-Oriented and Chitchat Dialogues: Proactive and Transition-Aware Conversational Agents (2025.emnlp-main)
Copied to clipboard
| Challenge: | Current efforts to bridge the two modes of interaction are reactive, focusing on responding to user inputs rather than coordinating dialogue flows. |
| Approach: | They propose a dataset designed for transition-aware dialogue modeling that incorporates structurally diverse and integrated mode flows. |
| Outcome: | The proposed dataset outperforms baseline models in intent detection and mode transition handling. |
Don’t Judge a Language Model by Its Last Layer: Contrastive Learning with Layer-Wise Attention Pooling (2022.coling-1)
Copied to clipboard
| Challenge: | Recent pre-trained language models (PLMs) have shown competitive performance on many natural language processing tasks. |
| Approach: | They propose a pooling strategy which preserves layer-wise signals captured in each layer and learns digested linguistic features for downstream tasks. |
| Outcome: | The proposed method improves on standard semantic textual similarity and semantic search tasks. |
PanicToCalm: A Proactive Counseling Agent for Panic Attacks (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing models for training such models are limited due to ethical and logistical issues. |
| Approach: | They propose a dataset that includes high-distress episodes constructed from first-person narratives and structured around the principles of Psychological First Aid. |
| Outcome: | The proposed model outperforms baseline models in counselor-side metrics and client affect improvement. |
How to Train Your Fact Verifier: Knowledge Transfer with Multimodal Open Models (2024.findings-emnlp)
Copied to clipboard
Jaeyoung Lee, Ximing Lu, Jack Hessel, Faeze Brahman, Youngjae Yu, Yonatan Bisk, Yejin Choi, Saadia Gabriel
| Challenge: | a growing influx of misinformation across news and social media is hampered by outdated foundation model training data. |
| Approach: | They propose to use large language models to scale up online policing mechanisms . they evaluate foundation model performance without continual updating . |
| Outcome: | The proposed model can improve performance without continual updating . the proposed model improves on two widely used benchmarks . |
Mitigating Framing Bias with Polarity Minimization Loss (2023.findings-emnlp)
Copied to clipboard
| Challenge: | polarity is a pervasive problem in modern media, misleading the understanding of what really happened via a skewed selection of information and language. |
| Approach: | They propose a loss function that encourages the model to minimize the polarity difference between the skewed input articles to reduce framing bias. |
| Outcome: | The proposed loss improves the model's ability to map polarity ends bidirectionally. |
Retrieval-Augmented Fine-Tuning With Preference Optimization For Visual Program Generation (2025.acl-long)
Copied to clipboard
| Challenge: | Visual programming languages (VPLs) allow users to create programs through graphical interfaces, which results in easier accessibility and widespread usage in various domains. |
| Approach: | They propose to train VPLs from user instructions using large language models . they propose to use retrieval-augmented fine-tuning to leverage repetitive use of subroutines . |
| Outcome: | The proposed method outperforms prompting-based methods for LD generation accuracy even with smaller backbone models. |
Progressive Facial Granularity Aggregation with Bilateral Attribute-based Enhancement for Face-to-Speech Synthesis (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for generating speech from facial images rely on pre-trained visual encoders and fine-tune them to align with speech embeddings. |
| Approach: | They propose to derive corresponding voices from facial images using face-to-voice synthesis, which derives corresponding voice from facial image. |
| Outcome: | The proposed approach significantly improves face-voice congruence and synthesis stability. |
Do Androids Laugh at Electric Sheep? Humor “Understanding” Benchmarks from The New Yorker Caption Contest (2023.acl-long)
Copied to clipboard
Jack Hessel, Ana Marasovic, Jena D. Hwang, Lillian Lee, Jeff Da, Rowan Zellers, Robert Mankoff, Yejin Choi
| Challenge: | Large neural networks can generate jokes, but do they really “understand” humor? a new challenge challenges AI models to match a joke to a cartoon, identify a winning caption, and explain why a winner is funny. |
| Approach: | They propose three tasks based on the New Yorker Cartoon Caption Contest . they aim to match a joke to a cartoon, identify a winning caption and explain why it's funny . |
| Outcome: | The proposed tasks are based on the New Yorker Cartoon Caption Contest . they include matching a joke to a cartoon, identifying a winning caption, and explaining why a funny caption is funny. |
Safeguarding RAG Pipelines with GMTP: A Gradient-based Masked Token Probability Method for Poisoned Document Detection (2025.findings-acl)
Copied to clipboard
| Challenge: | Retrieval-Augmented Generation (RAG) provides external knowledge for accurate and up-to-date responses, but external knowledge is vulnerable to poisoning and unauthorized injections. |
| Approach: | They propose a Gradient-based Masked Token Probability defense method to detect and filter out adversarially crafted documents by examining gradients of the retriever’s similarity function. |
| Outcome: | Experiments show that the proposed method eliminates over 90% of poisoned content while retaining relevant documents. |
Conversational Multi-Hop Reasoning with Neural Commonsense Knowledge and Symbolic Logic Rules (2021.emnlp-main)
Copied to clipboard
| Challenge: | Currently, conversational agents lack commonsense reasoning, preventing them from engaging in rich conversations with humans. |
| Approach: | They propose a commonsense reasoning system that uncovers unstated presumptions from user commands satisfying a general template of if-(state), then-(action), because-(goal) They propose to use a transformer-based generative commons sense knowledge base as its source of background knowledge to extract multi-hop reasoning chains from the neural KB. |
| Outcome: | The proposed model achieves a 35% higher success rate than existing methods with human users. |
Towards Few-shot Fact-Checking via Perplexity (2021.naacl-main)
Copied to clipboard
| Challenge: | Recent studies have shown that pre-trained language models can perform few-shot learning for various downstream tasks, such as question answering and machine translation. |
| Approach: | They propose a method to leverage the powerful transfer learning ability of a language model via a perplexity score to learn few-shot for the fact-checking task. |
| Outcome: | The proposed method outperforms the Major Class baseline by 10% on the F1-Macro metric across multiple datasets. |
Multi-Level Attention Aggregation for Language-Agnostic Speaker Replication (2024.eacl-short)
Copied to clipboard
| Challenge: | Recent advances in speech synthesis research have enabled the generation of natural-sounding speech, which has prompted a notable shift in TTS research towards the synthesis of speech in the voices of both seen and unseen speakers. |
| Approach: | They propose a multi-level attention aggregation approach that probes and amplifies various speaker-specific attributes in a hierarchical manner. |
| Outcome: | The proposed model achieves substantial speaker similarity and generalizes to out-of-domain (OOD) cases. |
Leveraging the Interplay between Syntactic and Acoustic Cues for Optimizing Korean TTS Pause Formation (2024.lrec-main)
Copied to clipboard
| Challenge: | despite recent advances in speech synthesis, the focus of research has been on high-resource languages like English. |
| Approach: | They propose a framework that incorporates modeling of syntactic and acoustic cues associated with pausing patterns. |
| Outcome: | The proposed framework generates natural speech even for longer and intricate out-of-domain sentences, despite training on short audio clips. |
Investigating Counterfactual Unfairness in LLMs towards Identities through Humor (2026.acl-long)
Copied to clipboard
Shubin Kim, Yejin Son, Junyeong Park, Keummin Ka, Seungbeen Lee, Jaeyoung Lee, Hyeju Jang, Alice Oh, Youngjae Yu
| Challenge: | Large Language Models (LLMs) absorb social and cultural biases embedded in vast web-scale corpora and are increasingly deployed in high-stakes domains such as hiring, education, and law. |
| Approach: | They propose a framework to investigate counterfactual unfairness through humor by observing how the model’s responses change when we swap who speaks and who is addressed while holding other factors constant. |
| Outcome: | The proposed framework covers humor generation refusal, speaker intention inference, and relational/societal impact prediction tasks. |
SQuARe: A Large-Scale Dataset of Sensitive Questions and Acceptable Responses Created through Human-Machine Collaboration (2023.acl-long)
Copied to clipboard
Hwaran Lee, Seokhee Hong, Joonsuk Park, Takyoung Kim, Meeyoung Cha, Yejin Choi, Byoungpil Kim, Gunhee Kim, Eun-Ju Lee, Yong Lim, Alice Oh, Sangchul Park, Jung-Woo Ha
| Challenge: | Existing studies focus on coping with social harms that large language models pose . however, discussions on sensitive issues can become toxic even if the users are well-intentioned. |
| Approach: | They propose to use Korean dataset to test whether LLMs can generate offensive content and propagate prejudices. |
| Outcome: | The proposed dataset shows that acceptable response generation improves for HyperCLOVA and GPT-3. |
Measuring Political Bias in Large Language Models: What Is Said and How It Is Said (2024.acl-long)
Copied to clipboard
| Challenge: | Existing benchmarks and measures focus on gender and racial biases, but political bias exists in LLMs and can lead to polarization and other harms in downstream applications. |
| Approach: | They propose to analyze the content and style of LLMs generated by political issues and propose a framework that can be scalable to other topics. |
| Outcome: | The proposed framework is easily scalable to other topics and is explainable. |
PicPersona-TOD : A Dataset for Personalizing Utterance Style in Task-Oriented Dialogue with Image Persona (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing systems produce generic, monotonic responses that lack individuality and fail to adapt to users’ personal attributes. |
| Approach: | They propose a dataset that incorporates user images as part of the persona, enabling personalized responses tailored to user-specific factors such as age or emotional context. |
| Outcome: | The proposed dataset enhances user experience, with personalized responses contributing to a more engaging interaction. |
RV-HATE: Reinforced Multi-Module Voting for Implicit Hate Speech Detection (2026.acl-long)
Copied to clipboard
| Challenge: | a new framework for hate speech detection addresses implicit hate speech by tailoring the detection process to dataset-specific attributes. |
| Approach: | They propose a framework to account for the dataset-specific characteristics of hate speech datasets. |
| Outcome: | The proposed framework improves detection accuracy and provides interpretable insights into the distinctive features of each dataset. |
BlendX: Complex Multi-Intent Detection with Blended Patterns (2024.lrec-main)
Copied to clipboard
| Challenge: | Existing datasets such as MixATIS and MixSNIPS have limitations in their formulation. |
| Approach: | They propose a set of multi-intent detection datasets that feature more diverse patterns than their predecessors. |
| Outcome: | The proposed datasets feature more diverse patterns than their predecessors and are more complex and diverse than existing datasets. |
AmpleHate: Amplifying the Attention for Versatile Implicit Hate Detection (2025.emnlp-main)
Copied to clipboard
| Challenge: | Current approaches to detect hate speech rely on contrastive learning to distinguish hate from non-hate sentences. |
| Approach: | They propose a novel approach to detect implicit hate speech by identifying explicit targets . they use a pretrained Named Entity Recognition model to capture explicit target information . |
| Outcome: | The proposed approach outperforms current methods and achieves faster convergence. |
Audio-Based Linguistic Feature Extraction for Enhancing Multi-lingual and Low-Resource Text-to-Speech (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods to synthesize speech for low-resource languages require a substantial amount of source language corpora to generate the linguistic knowledge that can be reused for speech synthesis. |
| Approach: | They propose a method that extracts linguistic features from audio input while effectively filtering out miscellaneous acoustic information including speaker-specific attributes like timbre. |
| Outcome: | The proposed method extracts linguistic features from audio input while effectively filtering out miscellaneous acoustic information including speaker-specific attributes like timbre. |
The BiGGen Bench: A Principled Benchmark for Fine-grained Evaluation of Language Models with Language Models (2025.naacl-long)
Copied to clipboard
Seungone Kim, Juyoung Suk, Ji Yong Cho, Shayne Longpre, Chaeeun Kim, Dongkeun Yoon, Guijin Son, Yejin Cho, Sheikh Shafayat, Jinheon Baek, Sue Hyun Park, Hyeonbin Hwang, Jinkyung Jo, Hyowon Cho, Haebin Shin, Seongyun Lee, Hanseok Oh, Noah Lee, Namgyu Ho, Se June Joo, Miyoung Ko, Yoonjoo Lee, Hyungjoo Chae, Jamin Shin, Joel Jang, Seonghyeon Ye, Bill Yuchen Lin, Sean Welleck, Graham Neubig, Moontae Lee, Kyungjae Lee, Minjoon Seo
| Challenge: | a recent study evaluated language models using abstract evaluation criteria that lack the flexibility and granularity of human assessment. |
| Approach: | They propose a benchmark to evaluate nine distinct language models' capabilities . they use instance-specific evaluation criteria to mirror human evaluation . |
| Outcome: | The proposed benchmark evaluates nine distinct capabilities of language models across 77 tasks. |
When One LLM Drools, Multi-LLM Collaboration Rules (2026.acl-long)
Copied to clipboard
Shangbin Feng, Wenxuan Ding, Alisa Liu, Zifeng Wang, Weijia Shi, Yike Wang, Shannon Zejiang Shen, Xiaochuang Han, Hunter Lang, Chen-Yu Lee, Tomas Pfister, Yejin Choi, Yulia Tsvetkov
| Challenge: | a single general-purpose LLM is not enough to produce a reliable output, argues this paper . a multi-LLM collaboration approach addresses reliability, democratization, and pluralism . |
| Approach: | They argue that a single general-purpose LLM is not enough to produce a reliable output . they organize existing multi-LLM collaboration methods into a hierarchy based on access and information exchange . |
| Outcome: | The proposed method addresses reliability, democratization, and pluralism challenges a single LLM fails to produce a reliable output. |
NeuS: Neutral Multi-News Summarization for Mitigating Framing Bias (2022.naacl-main)
Copied to clipboard
| Challenge: | a new task is proposed to reduce media news framing bias by generating a neutral summary from multiple news articles of the varying political leanings. |
| Approach: | They propose a task that generates a neutral summary from multiple news articles . they find title provides a good signal for framing bias and propose metric and model . |
| Outcome: | The proposed task can neutralize news content in hierarchical order from title to article . scalability remains a bottleneck due to the time-consuming human labor needed for composing the roundup . |
EnSToM: Enhancing Dialogue Systems with Entropy-Scaled Steering Vectors for Topic Maintenance (2025.findings-acl)
Copied to clipboard
| Challenge: | Small large language models (sLLMs) are lightweight and efficient, but struggle to maintain topic consistency in task-oriented dialogue systems. |
| Approach: | They propose an approach to ensure topic consistency in task-oriented dialogue systems by manipulating internal activations during inference. |
| Outcome: | The proposed approach achieves significant performance gain with a relatively small data size compared to fine-tuning approaches. |
MIRROR: Multimodal Cognitive Reframing Therapy for Rolling with Resistance (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent studies have explored the use of large language models (LLMs) in psychotherapy, however text-based cognitive behavioral therapy models struggle with client resistance, which weakens therapeutic alliance. |
| Approach: | They propose a multimodal approach that incorporates nonverbal cues and a synthetic dataset that pairs each client’s statements with corresponding facial images to train vision language models. |
| Outcome: | The proposed approach outperforms existing text-based cognitive behavioral therapy models in managing client resistance and fostering therapeutic alliance. |
Prompt-Guided Selective Masking Loss for Context-Aware Emotive Text-to-Speech (2025.findings-naacl)
Copied to clipboard
| Challenge: | Emotional dialogue speech synthesis (EDSS) aims to generate expressive speech by leveraging the dialogue context between interlocutors. |
| Approach: | They propose a large language model to generate holistic emotion tags based on prior dialogue context and pinpoint key words in the target utterance that align with the predicted emotional state. |
| Outcome: | The proposed method improves emotional expressiveness and facilitates automatic emotion speech generation during inference. |
Difficulty-Controllable Cloze Question Distractor Generation (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for generating high-quality distractors lack adaptability and control over difficulty levels. |
| Approach: | They propose a two-way distractor generation process to generate plausible distractors using an ensemble QA system and a multitask learning strategy to train a difficulty-controllable generation model. |
| Outcome: | The proposed method significantly outperforms GPT-4o in aligning distractor difficulty with human perception. |