Unverified50% confidenceFactExact time
Decoding adjustments do not touch model weights or internal representations, but instead modify the token selection strategy during the generation stage; contrastive decoding runs a strong model and a weak model simultaneously and takes the difference in their log probabilities as the score for the next token; DoLa compares the output distributions of different layers within a single model.
1
Sources
50%
Confidence
Long-term
Relevance
10/15/2026
First Seen
Sources
Related Entities
Related Claims
UnverifiedEach model invocation consumes both input tokens (prompts and context) and output tokens (the model's generated response), with the two typically priced separately.79% 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.76% similarUnverifiedThe core idea of contrastive decoding is to run a strong model and a weak model simultaneously and subtract their output logits to amplify the strong model's distinctive tendencies relative to the weak model.74% similarUnverifiedWhen fine-tuning embedding models, hard negatives (such as positions in the same domain but at different levels) should be used instead of random negatives, in order to learn more fine-grained semantic differences.73% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1038557API
curl https://kongchang.com/api/v1/knowledge/claims/1038557MCP
get_claim(id=1038557)