Dongsoo Lee

AnyBCQ: Hardware Efficient Flexible Binary-Coded Quantization for Multi-Precision LLMs

ICaRus: Identical Cache Reuse for Efficient Multi-Model Inference

SelfJudge: Faster Speculative Decoding via Self-Supervised Judge Verification

CodeGEMM: A Codebook-Centric Approach to Efficient GEMM in Quantized LLMs

FlexBCQ: Flexible Binary-coding Quantization for Large Language Models

Don’t Discard, but Keep It Small: Context-Preserving KV Cache Compression with Importance-Aware Adaptive Precision

To FP8 and Back Again: Quantifying Reduced Precision Effects on LLM Training Stability