SpecRouter dynamically schedules chains of draft and verifier models for multi-level speculative decoding, adding a mask-based rollback mechanism for key-value caches.
On the entropy of language models in getting semantic from tokens
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.LG 1years
2025 1verdicts
REJECT 1representative citing papers
citing papers explorer
-
SpecRouter: Adaptive Routing for Multi-Level Speculative Decoding in Large Language Models
SpecRouter dynamically schedules chains of draft and verifier models for multi-level speculative decoding, adding a mask-based rollback mechanism for key-value caches.