Yoon Kim

Log-Linear Attention

Learning to Interpret Weight Differences in Language Models

Distilling to Hybrid Attention Models via KL-Guided Layer Selection

Beyond Binary Rewards: Training LMs to Reason About Their Uncertainty

HybridCoT: Interleaving Latent and Text Chain-of-Thought for Efficient Reasoning

Self-Adapting Language Models

PaTH Attention: Position Encoding via Accumulating Householder Transformations

Training-Free Activation Sparsity in Large Language Models

The Semantic Hub Hypothesis: Language Models Share Semantic Representations Across Languages and Modalities

On the Duality between Gradient Transformations and Adapters

Ladder-Residual: Parallelism-Aware Architecture for Accelerating Large Model Inference with Communication Overlapping

Ladder Residual: Redefining Tensor Parallelism in Transformers for Accelerated Inference

The Surprising Effectiveness of Test-Time Training for Few-Shot Learning