publications

2025

1. General Reasoning Requires Learning to Reason from the Get-go

   Seungwook Han, Jyothish Pari, Samuel J. Gershman, and 1 more author

   2025

2024

1. Unveiling the Secret Recipe: A Guide For Supervised Fine-Tuning Small LLMs

   Aldo Pareja, Nikhil Shivakumar Nayak, Hao Wang, and 10 more authors

   2024
2. Emergence of Abstractions: Concept Encoding and Decoding Mechanism for In-Context Learning in Transformers

   Seungwook Han, Jinyeop Song, Jeff Gore, and 1 more author

   2024
3. Value Augmented Sampling for Language Model Alignment and Personalization

   Seungwook Han, Idan Shenfeld, Akash Srivastava, and 2 more authors

   2024

2023

1. Compositional Foundation Models for Hierarchical Planning

   Anurag Ajay, Seungwook Han, Yilun Du, and 7 more authors

   In Advances in Neural Information Processing Systems, 2023
2. Multi-Symmetry Ensembles: Improving Diversity and Generalization via Opposing Symmetries

   Charlotte Loh, Seungwook Han, Shivchander Sudalairaj, and 5 more authors

   In Proceedings of the 40th International Conference on Machine Learning, 23–29 jul 2023

   Abs PDF

   Deep ensembles (DE) have been successful in improving model performance by learning diverse members via the stochasticity of random initialization. While recent works have attempted to promote further diversity in DE via hyperparameters or regularizing loss functions, these methods primarily still rely on a stochastic approach to explore the hypothesis space. In this work, we present Multi-Symmetry Ensembles (MSE), a framework for constructing diverse ensembles by capturing the multiplicity of hypotheses along symmetry axes, which explore the hypothesis space beyond stochastic perturbations of model weights and hyperparameters. We leverage recent advances in contrastive representation learning to create models that separately capture opposing hypotheses of invariant and equivariant functional classes and present a simple ensembling approach to efficiently combine appropriate hypotheses for a given task. We show that MSE effectively captures the multiplicity of conflicting hypotheses that is often required in large, diverse datasets like ImageNet. As a result of their inherent diversity, MSE improves classification performance, uncertainty quantification, and generalization across a series of transfer tasks. Our code is available at https://github.com/clott3/multi-sym-ensem

3. Estimating the Density Ratio between Distributions with High Discrepancy using Multinomial Logistic Regression

   Akash Srivastava, Seungwook Han, Kai Xu, and 2 more authors

   23–29 jul 2023
4. Constructive Assimilation: Boosting Contrastive Learning Performance through View Generation Strategies

   Ligong Han, Seungwook Han, Shivchander Sudalairaj, and 8 more authors

   23–29 jul 2023

2022

1. On the Importance of Calibration in Semi-supervised Learning

   Charlotte Loh, Rumen Dangovski, Shivchander Sudalairaj, and 5 more authors

   23–29 jul 2022
2. Equivariant Contrastive Learning

   Rumen Dangovski, Li Jing, Charlotte Loh, and 5 more authors

   23–29 jul 2022

2021

1. Gage MPC: Bypassing Residual Function Leakage for Non-Interactive MPC

   Ghada Almashaqbeh, Fabrice Benhamouda, Seungwook Han, and 6 more authors

   23–29 jul 2021

2020

1. 3D distributed deep learning framework for prediction of human intelligence from brain MRI

   Seungwook Han, Yan Zhang, Yihui Ren, and 3 more authors

   In Medical Imaging 2020: Biomedical Applications in Molecular, Structural, and Functional Imaging, 23–29 jul 2020

   DOI
2. not-so-BigGAN: Generating High-Fidelity Images on Small Compute with Wavelet-based Super-Resolution

   Seungwook Han, Akash Srivastava, Cole Hurwitz, and 2 more authors

   23–29 jul 2020

2019

1. Predicting the accuracy of neural networks from final and intermediate layer outputs

   Chad DeChant, Seungwook Han, and Hod Lipson

   In , 23–29 jul 2019