Nightjar: Dynamic adaptive speculative decoding for large language models serving
Authors:
- Rui LiRLRui Li
- Zhaoning ZhangZZZhaoning Zhang
- Libo ZhangLZLibo Zhang
- Huaimin WangHWHuaimin Wang
- Xiang FuXFXiang Fu
- Zhiquan LaiZLZhiquan Lai
- 12 June 2026
- journal article
- Published by Elsevier in Journal of Systems Architecture
- Vol. 178
Abstract
No abstract availableFunding Information
- National Natural Science Foundation of China (62421002)
This publication has 8 references indexed in Scilit:
- ViDA: Lossless VideoQA Acceleration via Selective Sparse Self-Speculation With Parallel Computational Load ManagementIEEE Transactions on Parallel and Distributed Systems, 2026
- Oneiros: KV Cache Optimization through Parameter Remapping for Multi-tenant LLM Serving2025
- TightLLM: Maximizing Throughput for LLM Inference via Adaptive Offloading PolicyIEEE Transactions on Computers, 2025
- OPT-Tree: Speculative Decoding with Adaptive Draft Tree StructureTransactions of the Association for Computational Linguistics, 2025
- Dovetail: A CPU/GPU Heterogeneous Speculative Decoding for LLM inference2025
- Efficient Memory Management for Large Language Model Serving with PagedAttention2023
- A survey of techniques for optimizing transformer inferenceJournal of Systems Architecture, 2023
- A contextual-bandit approach to personalized news article recommendation2010