Imagen de Portada

Training-free Dropout Sampling for Semantic Token Acceptance in Speculative Decoding

Fuente: arXiv

Guardado en:

Detalles Bibliográficos
Autores principales:	Lee, Jeongtae, Jo, Minjung, Jeong, Hyunjoon, Park, Gunho, Woo, Sunghyeon, Kim, Joonghoon, Kwon, Se Jung, Lee, Dongsoo
Formato:	Preprint
Publicado:	2026
Materias:	Computation and Language
Acceso en línea:	Acceder al recurso
Etiquetas:	Agregar Etiqueta Sin Etiquetas, Sea el primero en etiquetar este registro!

Ejemplares similares

PrefillShare: A Shared Prefill Module for KV Reuse in Multi-LLM Disaggregated Serving
por: Woo, Sunghyeon, et al.
Publicado: (2026)

DropBP: Accelerating Fine-Tuning of Large Language Models by Dropping Backward Propagation
por: Woo, Sunghyeon, et al.
Publicado: (2024)

SelfJudge: Faster Speculative Decoding via Self-Supervised Judge Verification
por: Yoon, Kanghoon, et al.
Publicado: (2025)

ICaRus: Identical Cache Reuse for Efficient Multi Model Inference
por: Woo, Sunghyeon, et al.
Publicado: (2026)

SUN: Shared Use of Next-token Prediction for Efficient Multi-LLM Disaggregated Serving
por: Woo, Sunghyeon, et al.
Publicado: (2026)

Affine-Scaled Attention: Towards Flexible and Stable Transformer Attention
por: Bae, Jeongin, et al.
Publicado: (2026)

AnyBCQ: Hardware Efficient Flexible Binary-Coded Quantization for Multi-Precision LLMs
por: Park, Gunho, et al.
Publicado: (2025)

An Inquiry into Datacenter TCO for LLM Inference with FP8
por: Kim, Jiwoo, et al.
Publicado: (2025)

No Token Left Behind: Reliable KV Cache Compression via Importance-Aware Mixed Precision Quantization
por: Yang, June Yong, et al.
Publicado: (2024)

FlexRound: Learnable Rounding based on Element-wise Division for Post-Training Quantization
por: Lee, Jung Hyun, et al.
Publicado: (2023)

LUT-GEMM: Quantized Matrix Multiplication based on LUTs for Efficient Inference in Large-Scale Generative Language Models
por: Park, Gunho, et al.
Publicado: (2022)

CodeGEMM: A Codebook-Centric Approach to Efficient GEMM in Quantized LLMs
por: Park, Gunho, et al.
Publicado: (2025)

To FP8 and Back Again: Quantifying Reduced Precision Effects on LLM Training Stability
por: Lee, Joonhyung, et al.
Publicado: (2024)

FIGLUT: An Energy-Efficient Accelerator Design for FP-INT GEMM Using Look-Up Tables
por: Park, Gunho, et al.
Publicado: (2025)

Rethinking Channel Dimensions to Isolate Outliers for Low-bit Weight Quantization of Large Language Models
por: Heo, Jung Hwan, et al.
Publicado: (2023)

Multi-Drafter Speculative Decoding with Alignment Feedback
por: Kim, Taehyeon, et al.
Publicado: (2026)

LRQ: Optimizing Post-Training Quantization for Large Language Models by Learning Low-Rank Weight-Scaling Matrices
por: Lee, Jung Hyun, et al.
Publicado: (2024)

Unifying Uniform and Binary-coding Quantization for Accurate Compression of Large Language Models
por: Park, Seungcheol, et al.
Publicado: (2025)

Faster Inference of LLMs using FP8 on the Intel Gaudi
por: Lee, Joonhyung, et al.
Publicado: (2025)

3-Model Speculative Decoding
por: Byun, Sanghyun, et al.
Publicado: (2025)

Cactus: Accelerating Auto-Regressive Decoding with Constrained Acceptance Speculative Sampling
por: Hao, Yongchang, et al.
Publicado: (2026)

PaCA: Partial Connection Adaptation for Efficient Fine-Tuning
por: Woo, Sunghyeon, et al.
Publicado: (2025)

AdaEDL: Early Draft Stopping for Speculative Decoding of Large Language Models via an Entropy-based Lower Bound on Token Acceptance Probability
por: Agrawal, Sudhanshu, et al.
Publicado: (2024)

ACoRN: Noise-Robust Abstractive Compression in Retrieval-Augmented Language Models
por: Kim, Singon, et al.
Publicado: (2025)

AM-SORT: Adaptable Motion Predictor with Historical Trajectory Embedding for Multi-Object Tracking
por: Kim, Vitaliy, et al.
Publicado: (2024)

Accelerating Codec-based Speech Synthesis with Multi-Token Prediction and Speculative Decoding
por: Nguyen, Tan Dat, et al.
Publicado: (2024)

RaDL: Relation-aware Disentangled Learning for Multi-Instance Text-to-Image Generation
por: Park, Geon, et al.
Publicado: (2025)

Insertional Mutations in RsFLC2 Disrupt Vernalization‐Induced Intragenic Looping and H3K27me3 Enrichment in Radish
por: Sang Woo Lee, et al.
Publicado: (2025)

Let Triggers Control: Frequency-Aware Dropout for Effective Token Control
por: Koh, Junyoung, et al.
Publicado: (2026)

BudgetDraft: Acceptance-Aware Multi-View Training for Sparse-KV Speculative Decoding
por: He, Liang, et al.
Publicado: (2026)

Principled Coarse-Grained Acceptance for Speculative Decoding in Speech
por: Yanuka, Moran, et al.
Publicado: (2025)

Acceptance Dynamics Across Cognitive Domains in Speculative Decoding
por: Mahmoud, Saif
Publicado: (2026)

Speculative Verification: Exploiting Information Gain to Refine Speculative Decoding
por: Kim, Sungkyun, et al.
Publicado: (2025)

Lacunary elliptic maximal operator on the Heisenberg group
por: Kim, Joonil, et al.
Publicado: (2025)

Recursive Speculative Decoding: Accelerating LLM Inference via Sampling Without Replacement
por: Jeon, Wonseok, et al.
Publicado: (2024)

Training-free Mixed-Resolution Latent Upsampling for Spatially Accelerated Diffusion Transformers
por: Jeong, Wongi, et al.
Publicado: (2025)

Effective Field Theory of a Noncollinear Altermagnet
por: Lee, Seungho, et al.
Publicado: (2025)

Lossless Acceleration of Large Language Models with Hierarchical Drafting based on Temporal Locality in Speculative Decoding
por: Cho, Sukmin, et al.
Publicado: (2025)

LK Losses: Direct Acceptance Rate Optimization for Speculative Decoding
por: Samarin, Alexander, et al.
Publicado: (2026)

Text-guided Weakly Supervised Framework for Dynamic Facial Expression Recognition
por: Jung, Gunho, et al.
Publicado: (2025)