Self-Speculative Decoding
A speculative decoding scheme that uses flow-based drafts for autoregressive verification within a single model, requiring no separate draft model and substantially improving throughput.
Timeline (last 90 days)
Lafa designs a self-speculative decoding scheme that uses flow-based drafts for autoregressive verification within a single model, requiring no additional draft model parameters or memory overhead.
At comparable quality, Lafa's self-speculative decoding scheme achieves 2.54x the throughput of autoregressive models, outperforming multi-token prediction and discrete diffusion methods.
All Facts (2)
Lafa designs a self-speculative decoding scheme that uses flow-based drafts for autoregressive verification within a single model, requiring no additional draft model parameters or memory overhead.
50%UnverifiedAt comparable quality, Lafa's self-speculative decoding scheme achieves 2.54x the throughput of autoregressive models, outperforming multi-token prediction and discrete diffusion methods.
50%