by Jonathan Mamou Papers
2 papers found
Accelerating LLM Inference with Lossless Speculative Decoding Algorithms for Heterogeneous Vocabularies
Nadav Timor, Jonathan Mamou, Daniel Korat et al.
ICML 2025oral
9
citations
Distributed Speculative Inference (DSI): Speculation Parallelism for Provably Faster Lossless Language Model Inference
Nadav Timor, Jonathan Mamou, Daniel Korat et al.
ICLR 2025posterarXiv:2405.14105
7
citations