Publications

[C9] PertReason: A Knowledge-Grounded Benchmark and Framework for Cell-State-Conditioned Mechanistic Reasoning of Perturbation Effects

Dongkwan Kim*, Yiming Gao*, Yining Yang, Yang Shen. Neural Information Processing Systems Evaluations & Datasets Track (NeurIPS E&D), 2026

One-sentence Summary: We propose PertReason, a knowledge-grounded QA benchmark designed to evaluate how models mechanistically reason through cell-state-specific perturbation effects by leveraging causal pathways.

arXiv

[C8] MUG-Eval: A Proxy Evaluation Framework for Multilingual Generation Capabilities in Any Language

Seyoung Song, Seogyeong Jeong, Eunsu Kim, Jiho Jin, Dongkwan Kim, Jay Shin, Alice Oh. Findings of the Empirical Methods in Natural Language Processing: EMNLP 2025 (EMNLP-Findings), 2025

One-sentence Summary: A language-agnostic framework that evaluates LLM multilingual generation by measuring task completion rates in self-communication scenarios, enabling objective assessment across 2,100+ languages without requiring language-specific tools or humans.

Paper arXiv Code Poster

[C7] LLM-as-an-Interviewer: Beyond Static Testing Through Dynamic LLM Evaluation

Eunsu Kim, Juyoung Suk, Seungone Kim, Niklas Muennighoff, Dongkwan Kim, Alice Oh. Findings of the Association for Computational Linguistics: ACL 2025 (ACL-Findings), 2025

One-sentence Summary: We propose a new paradigm for evaluating large language models (LLMs), called LLM-as-an-Interviewer.

Paper arXiv Code Poster

[C6] Generalizing Weisfeiler-Lehman Kernels to Subgraphs

Dongkwan Kim and Alice Oh. International Conference on Learning Representations (ICLR), 2025

One-sentence Summary: We propose WLKS that extends the WL graph kernel to subgraphs, captures high-order structural similarities in k-hop neighborhoods, efficiently outperforming state-of-the-art GNNs.

Paper arXiv Code Poster

[C5] Perceptions to Beliefs: Exploring Precursory Inferences for Theory of Mind in Large Language Models

Chani Jung, Dongkwan Kim, Jiho Jin, Jiseon Kim, Yeon Seonwoo, Yejin Choi, Alice Oh, and Hyunwoo Kim. Empirical Methods in Natural Language Processing (EMNLP), 2024

One-sentence Summary: We assess key precursors of Theory of Mind (ToM) in LLMs by perception-augmented ToM benchmarks. We propose PercepToM, a ToM method inspired by our findings of models’ strength in perception inference and weakness in perception-to-belief inference.

Paper arXiv Code Data Poster

[C4] Translating Subgraphs to Nodes Makes Simple GNNs Strong and Efficient for Subgraph Representation Learning

Dongkwan Kim and Alice Oh. International Conference on Machine Learning (ICML), 2024

One-sentence Summary: We propose Subgraph-To-Node translation to effectively and efficiently learn representations of subgraphs by coarsely translating subgraphs into nodes.

Paper arXiv Code Poster

[C3] Models and Benchmarks for Representation Learning of Partially Observed Subgraphs

Dongkwan Kim, Jiho Jin, Jaimeen Ahn and Alice Oh. International Conference on Information and Knowledge Management (CIKM), 2022

Paper arXiv Code

[C2] How to Find Your Friendly Neighborhood: Graph Attention Design with Self-Supervision

Dongkwan Kim and Alice Oh. International Conference on Learning Representations (ICLR), 2021

One-sentence Summary: We propose a method that self-supervise graph attention through edges and it should be designed according to the average degree and homophily of graphs.

Paper arXiv Code Poster Slide

[C1] Homogeneity-Based Transmissive Process To Model True and False News in Social Networks

Jooyeon Kim, Dongkwan Kim, and Alice Oh. International Conference on Web Search and Data Mining (WSDM), 2019

Paper arXiv Code Data

[P2] PertReason: A Knowledge-Grounded Benchmark and Framework for Cell-State-Conditioned Mechanistic Reasoning of Perturbation Effects

Dongkwan Kim, Yiming Gao, Yining Yang, Yang Shen. Arxiv, 2026

arXiv

[P1] Measuring Interest Group Positions on Legislation: An AI-Driven Analysis of Lobbying Reports

Jiseon Kim, Dongkwan Kim, Joohye Jeong, Alice Oh, In Song Kim. Arxiv, 2025

arXiv Code

[W6] PertReasonQA: A Knowledge-Grounded Benchmark and Framework for Cell-State–Conditioned Mechanistic Reasoning of Perturbation Effects

Dongkwan Kim, Yiming Gao, Yining Yang, Yang Shen. Workshop on Multi-modal Foundation Models and Large Language Models for Life Sciences at ICML (ICML FM4LS), 2026

One-sentence Summary: We propose PertReasonQA, a knowledge-grounded QA benchmark designed to evaluate how models mechanistically reason through cell-state-specific perturbation effects by leveraging causal pathways.

Paper

[W5] Salad-Bowl-LLM: Multi-Culture LLMs by In-Context Demonstrations from Diverse Cultures

Dongkwan Kim, Junho Myung, and Alice Oh. Workshop on Socially Responsible Language Modelling Research at NeurIPS (NeurIPS SoLaR), 2024

One-sentence Summary: We explore the use of in-context learning with diverse demonstrations to enhance Large Language Models’ cultural understanding.

Paper Poster

[W4] Understanding Lobbying Strategies in Legislative Process: Bill Position Dataset and Lobbying Analysis

Jiseon Kim, Dongkwan Kim, Joohye Jeong, In Song Kim, Alice Oh. Workshop for Women in Machine Learning at NeurIPS (NeurIPS WiML), 2024

Poster

[W3] Efficient Representation Learning of Subgraphs by Subgraph-To-Node Translation

Dongkwan Kim and Alice Oh. Workshop on Geometrical and Topological Representation Learning at ICLR (ICLR GTRL), 2022

Note: The extended version is published at ICML 2024.

Paper arXiv

[W2] Supervised Graph Attention Network for Semi-Supervised Node Classification

Dongkwan Kim and Alice Oh. Workshop on Graph Representation Learning at NeurIPS (NeurIPS GRL), 2019

Note: The full version is published at ICLR 2021.

Paper

[W1] Additive Compositionality of Word Vectors

Yeon Seonwoo, Sungjoon Park, Dongkwan Kim, and Alice Oh. Workshop on Noisy User-generated Text at EMNLP (EMNLP W-NUT), 2019

Paper