Unverified50% confidenceFactExact time
The losslessness of speculative decoding relies on a modified rejection sampling mechanism: for each token proposed by the draft model, the target model computes its probability and decides acceptance or rejection accordingly; once rejected, resampling is done from the difference between the target model and draft model distributions (after normalization).
1
Sources
50%
Confidence
Long-term
Relevance
10/20/2026
First Seen
Sources
Related Entities
Related Claims
VerifiedSpeculative decoding generates candidate tokens through a lightweight draft model, with the main model verifying them in parallel in a single forward pass, accepting tokens that conform to the distribution and regenerating from the first rejected position.80% similarUnverifiedAdaptiveSpec's progressive marginal rule allows a draft token to be accepted when the ratio of the target model's probability on that draft token to its top-1 token probability exceeds a threshold, even if the two do not exactly match.77% similarVerifiedSpeculative decoding is a lossless acceleration technique that does not alter the model's output distribution; it uses a draft model with far fewer parameters than the target model to generate candidate tokens, which the target model then verifies in parallel.75% similarUnverifiedSpeculative sampling uses a cheap draft model to generate multiple candidate states in parallel, which are then verified in parallel by the target model, reducing the number of target model calls while keeping the target distribution unchanged.73% similarUnverifiedA model's refusal behavior is a probabilistic response shaped during training, rather than a response based on genuine understanding or judgment.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1074370API
curl https://kongchang.com/api/v1/knowledge/claims/1074370MCP
get_claim(id=1074370)