BubbleSpec exploits long-tail bubbles in synchronous RL by using faster ranks' idle time to pre-generate rollout drafts for speculative decoding, reducing steps by 50% and raising throughput up to 1.8x while preserving exact synchrony.
Jones, Zheng Dong, and Peipei Zhou
2 Pith papers cite this work. Polarity classification is still indexing.
2
Pith papers citing it
years
2026 2verdicts
UNVERDICTED 2representative citing papers
Systematic study concludes overlay architectures suit frequent model switching in current autonomous driving setups, while customized ones may become preferable as bitstream reload overhead decreases.
citing papers explorer
-
BubbleSpec: Turning Long-Tail Bubbles into Speculative Rollout Drafts for Synchronous Reinforcement Learning
BubbleSpec exploits long-tail bubbles in synchronous RL by using faster ranks' idle time to pre-generate rollout drafts for speculative decoding, reducing steps by 50% and raising throughput up to 1.8x while preserving exact synchrony.
-
To Overlay or to Customize? Revisiting Architectural Choices in Heterogeneous Systems
Systematic study concludes overlay architectures suit frequent model switching in current autonomous driving setups, while customized ones may become preferable as bitstream reload overhead decreases.