Papers by Seungju Han
Subtle Risks, Critical Failures: A Framework for Diagnosing Physical Safety of LLMs for Embodied Decision Making (2025.emnlp-main)
Copied to clipboard
Yejin Son, Minseo Kim, Sungwoong Kim, Seungju Han, Jian Kim, Dongju Jang, Youngjae Yu, Chan Young Park
| Challenge: | Existing safety evaluations rely on coarse success rates and domain-specific setups, making it difficult to diagnose why and where these models fail. |
| Approach: | They propose a framework for systematically evaluating the physical safety of LLMs in embodied decision making. |
| Outcome: | The proposed framework assesses the physical safety of LLMs in embodied decision making. |
SMILE: Multimodal Dataset for Understanding Laughter in Video with Language Models (2024.findings-naacl)
Copied to clipboard
| Challenge: | Despite advances in artificial intelligence, building social intelligence remains a challenge. |
| Approach: | They propose a task to explain why people laugh in a video and a dataset to do this. |
| Outcome: | The proposed dataset generates plausible explanations for laughter in video and in-the-wild videos. |
Representation Bending for Large Language Model Safety (2025.acl-long)
Copied to clipboard
Ashkan Yousefpour, Taeheon Kim, Ryan Sungmo Kwon, Seungbeen Lee, Wonje Jeung, Seungju Han, Alvin Wan, Harrison Ngan, Youngjae Yu, Jonghyun Choi
| Challenge: | Existing safety-enhancing techniques, such as fine-tuning with human feedback or adversarial training, are still vulnerable as they address specific threats and fail to generalize across unseen attacks. |
| Approach: | They propose a new approach that disrupts representations underlying harmful behaviors in Large Language Models by using loss-based fine-tuning. |
| Outcome: | The proposed approach outperforms existing methods such as Circuit Breaker, RMU, and NPO with 95% reduction in attack success rates across diverse jailbreak benchmarks. |
Selective Vision is the Challenge for Visual Reasoning: A Benchmark for Visual Argument Understanding (2024.emnlp-main)
Copied to clipboard
| Challenge: | Visual arguments rely on images to persuade viewers to do or believe something . |
| Approach: | They propose three tasks for evaluating visual argument understanding . they use visual premises, commonsense premises and reasoning trees to analyze visual arguments . |
| Outcome: | The proposed tasks evaluate visual argument understanding using a dataset of 1,611 images annotated with 5,112 visual premises (with regions), 5,574 commonsense premises, and reasoning trees connecting them into structured arguments. |
Nemotron-CrossThink: Scaling Self-Learning beyond Math Reasoning (2026.eacl-long)
Copied to clipboard
Syeda Nahida Akter, Shrimai Prabhumoye, Matvei Novikov, Seungju Han, Ying Lin, Evelina Bakhturina, Eric Nyberg, Yejin Choi, Mostofa Patwary, Mohammad Shoeybi, Bryan Catanzaro
| Challenge: | Prior work has successfully applied Reinforcement Learning (RL) to mathematical reasoning, but generalization to broader domains remains challenging due to limited data and lack of verifiable rewards for unstructured domains. |
| Approach: | They propose a framework that integrates multi-domain corpora into RL training to improve generalization across diverse reasoning tasks. |
| Outcome: | The proposed framework improves generalization across diverse reasoning tasks. |
Reading Books is Great, But Not if You Are Driving! Visually Grounded Reasoning about Defeasible Commonsense Norms (2023.emnlp-main)
Copied to clipboard
Seungju Han, Junhyeok Kim, Jack Hessel, Liwei Jiang, Jiwan Chung, Yejin Son, Yejin Choi, Youngjae Yu
| Challenge: | NormLens is a visual-grounded framework for understanding commonsense norms . state-of-the-art models are not well-aligned with human annotation, we show . |
| Approach: | They propose a visual-grounded framework to study commonsense norms by NormLens . they find that models are not well-aligned with human annotation . |
| Outcome: | The proposed model judgments and explanations are not well-aligned with human annotations. |
Meet Your Favorite Character: Open-domain Chatbot Mimicking Fictional Characters with only a Few Utterances (2022.naacl-main)
Copied to clipboard
| Challenge: | In this paper, we consider mimicking fictional characters as a promising direction for building engaging conversation models. |
| Approach: | They propose a task where only a few utterances of each fictional character are available to generate responses mimicking them. |
| Outcome: | The proposed method generates responses better reflecting the style of fictional characters than baseline methods. |
Measuring and Improving Semantic Diversity of Dialogue Generation (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Existing evaluation metrics for response diversity do not capture the semantic diversity of generated responses. |
| Approach: | They propose an automatic evaluation metric to measure the semantic diversity of generated responses . they show that it captures human judgments better than existing diversity metrics . |
| Outcome: | The proposed metric captures human judgments on response diversity better than existing lexical diversity metrics. |
MAPoRL: Multi-Agent Post-Co-Training for Collaborative Large Language Models with Reinforcement Learning (2025.acl-long)
Copied to clipboard
| Challenge: | Existing studies focus on prompting and developing workflows with frozen LLMs. |
| Approach: | They propose a multi-agentic framework for collaborative LLMs with reinforcement learning that leverages multi-gendered frameworks to enhance collaboration. |
| Outcome: | The proposed model improves collaboration performance across multiple datasets with generalization to unseen domains compared to existing models. |
Do LLMs Have Distinct and Consistent Personality? TRAIT: Personality Testset designed for LLMs with Psychometrics (2025.findings-naacl)
Copied to clipboard
Seungbeen Lee, Seungwon Lim, Seungju Han, Giyeong Oh, Hyungjoo Chae, Jiwan Chung, Minju Kim, Beong-woo Kwak, Yeonsoo Lee, Dongha Lee, Jinyoung Yeo, Youngjae Yu
| Challenge: | Recent advances in Large Language Models (LLMs) have led to their adaptation as conversational agents. |
| Approach: | They propose a new benchmark that uses 8K multi-choice questions to assess the personality of Large Language Models. |
| Outcome: | The proposed personality test outperforms existing personality tests for LLMs in reliability and validity. |
DUSK: Do Not Unlearn Shared Knowledge (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent work suggests that machine learning models are indistinguishable from models trained on retain sets. |
| Approach: | They propose a benchmark to evaluate machine unlearning under realistic knowledge overlap . they construct documents containing both shared and unique knowledge . |
| Outcome: | The proposed model is indistinguishable from a model retrained on the retain set while only forget-specific content is removed. |
Distilling the Knowledge of Large-scale Generative Models into Retrieval Models for Efficient Open-domain Conversation (2021.findings-emnlp)
Copied to clipboard
| Challenge: | generative models are less practical for building real-time conversation systems due to high latency and large memory footprint. |
| Approach: | They propose a method that preserves the efficiency of a retrieval model while leveraging the conversational ability of generative models. |
| Outcome: | The proposed method preserves the efficiency of a retrieval model while leveraging the conversational ability of generative models. |