Pang Wei Koh

DataDecide: How to Predict Best Pretraining Data with Small Experiments ICML 2025 Exploring How Generative MLLMs Perceive More Than CLIP with the Same Vision Encoder ACL 2025 Group-robust Sample Reweighting for Subpopulation Shifts via Influence Functions ICLR 2025 Language models scale reliably with over-training and on downstream tasks ICLR 2025 NICE Data Selection for Instruction Tuning in LLMs with Non-differentiable Evaluation Metric ICML 2025 OLMoE: Open Mixture-of-Experts Language Models ICLR 2025 OLMoTrace: Tracing Language Model Outputs Back to Trillions of Training Tokens ACL 2025 PLeaS - Merging Models with Permutations and Least Squares CVPR 2025 S4S: Solving for a Fast Diffusion Model Solver ICML 2025 Annotation alignment: Comparing LLM and human annotations of conversational safety EMNLP 2024 CopyBench: Measuring Literal and Non-Literal Reproduction of Copyright-Protected Text in Language Model Generation EMNLP 2024 DataComp-LM: In search of the next generation of training sets for language models NeurIPS 2024 Improving Domain Generalization with Domain Relations ICLR 2024 Instructional Fingerprinting of Large Language Models NAACL 2024 MediQ: Question-Asking LLMs and a Benchmark for Reliable Interactive Clinical Reasoning NeurIPS 2024 Merge to Learn: Efficiently Adding Skills to Language Models with Model Merging EMNLP 2024 Multilingual Diversity Improves Vision-Language Representations NeurIPS 2024 Position Paper: Data-Centric AI in the Age of Large Language Models EMNLP 2024 Scaling Retrieval-Based Language Models with a Trillion-Token Datastore NeurIPS 2024 The Generative AI Paradox: “What It Can Create, It May Not Understand” ICLR 2024 The Unmet Promise of Synthetic Training Images: Using Retrieved Real Images Performs Better NeurIPS 2024 Uncertainty of Thoughts: Uncertainty-Aware Planning Enhances Information Seeking in LLMs NeurIPS 2024 Out-of-Domain Robustness via Targeted Augmentations ICML 2023 Extending the WILDS Benchmark for Unsupervised Adaptation ICLR 2022 Accuracy on the Line: on the Strong Correlation Between Out-of-Distribution and In-Distribution Generalization ICML 2021 Just Train Twice: Improving Group Robustness without Training Group Information ICML 2021 Selective Classification Can Magnify Disparities Across Groups ICLR 2021 WILDS: A Benchmark of in-the-Wild Distribution Shifts ICML 2021 An Investigation of Why Overparameterization Exacerbates Spurious Correlations ICML 2020 Concept Bottleneck Models ICML 2020 ExpBERT: Representation Engineering with Natural Language Explanations ACL 2020 Inferring Multidimensional Rates of Aging from Cross-Sectional Data AISTATS 2019 Understanding Black-box Predictions via Influence Functions ICML 2017