Youngsuk Park

Research@Amazon, AWS AI Labs·美国·OpenReview

研究方向

inference optimization · large language model · efficient training · memory efficient optimization · optimization · time series · forecast · reinforcement learning · graphical model

6.5

Not-a-Bandit: Provably No-Regret Drafter Selection in Speculative Decoding for LLMs

ICLR 2026Poster

6.1

Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization

ICML 2025Poster

5.5

PROXSPARSE: REGULARIZED LEARNING OF SEMI-STRUCTURED SPARSITY MASKS FOR PRETRAINED LLMS

ICML 2025Poster

5.5

RoSTE: An Efficient Quantization-Aware Supervised Fine-Tuning Approach for Large Language Models

ICML 2025Poster

5.5

Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization

合作者 (20)

Youngsuk Park

Scaling Laws Meet Model Architecture: Toward Inference-Efficient LLMs

MuonBP: Faster Muon via Block-Periodic Orthogonalization

TritonRL: Training LLMs to Think and Code Triton Without Cheating

Not-a-Bandit: Provably No-Regret Drafter Selection in Speculative Decoding for LLMs

Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization

PROXSPARSE: REGULARIZED LEARNING OF SEMI-STRUCTURED SPARSITY MASKS FOR PRETRAINED LLMS

RoSTE: An Efficient Quantization-Aware Supervised Fine-Tuning Approach for Large Language Models

Enhancing Foundation Models for Time Series Forecasting via Wavelet-based Tokenization