Kyunghyun Cho

Temporal Generalization: A Reality Check

On the Relationship Between the Choice of Representation and In-Context Learning

Multi-modal Data Spectrum: Multi-modal Datasets are Multi-dimensional

Forget Forgetting: Continual Learning in a World of Abundant Memory

Reference-Specific Unlearning Metrics Can Hide the Truth: A Reality Check

scCBGM: Single-Cell Editing via Concept Bottlenecks

None to Optima in Few Shots: Bayesian Optimization with MDP Priors

Why Alignment Must Precede Distillation: A Minimal Working Explanation

Group Contrastive Learning for Weakly Paired Multimodal Data

EVA-Flow: Environment-Aware Flow Matching for Unified 3D Molecular Conformation Generation

Test Time Scaling for Neural Processes

Predicting partially observable dynamical systems via diffusion models with a multiscale inference scheme

Why Knowledge Distillation Works in Generative Models: A Minimal Working Explanation

AION-1: Omnimodal Foundation Model for Astronomical Sciences

Efficient semantic uncertainty quantification in language models via diversity-steered sampling

Concept Bottleneck Language Models For Protein Design

Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models

Aioli: A Unified Optimization Framework for Language Model Data Mixing

$\mathbb{X}$-Sample Contrastive Loss: Improving Contrastive Learning with Sample Similarity Graphs

Implicit Generative Property Enhancer

Generating All-Atom Protein Structure from Sequence-Only Training Data

Generalists vs. Specialists: Evaluating LLMs on Highly-Constrained Biophysical Sequence Optimization Tasks

Supervised Contrastive Block Disentanglement

Hyperparameter Loss Surfaces Are Simple Near their Optima

Hyperparameters in Continual Learning: A Reality Check

Generalizing to any diverse distribution: uniformity, gentle finetuning & rebalancing

Molecular Active Learning: How can LLMs Help?

An Asymptotic Theory of Random Search for Hyperparameters in Deep Learning