conftrace_

candidate tokens

2 papers

Explore in graph

Co-occurring keywords · papers mentioning both

generation process (1) test-time overhead (1) values (1) cached keys (1) bradley-terry loss (1) confidence score (1) draft model (1) speculative decoding (1) reward model (1) language model (1)

Papers

Towards Cost-Effective Reward Guided Text Generation ICML 2025 GliDe with a CaPE: A Low-Hassle Method to Accelerate Speculative Decoding ICML 2024