Adding confidence-gated repair branches to block-parallel drafts increases accepted token count slightly but, in the reported setup, reduces end-to-end speedup relative to the repair-free parallel baseline.
Advances in Neural Information Processing Systems , year=
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.AI 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
CURE: Local Uncertainty Repair for Block-Parallel Speculative Decoding
Adding confidence-gated repair branches to block-parallel drafts increases accepted token count slightly but, in the reported setup, reduces end-to-end speedup relative to the repair-free parallel baseline.