Unverified50% confidenceFactExact time
There is a MAP decoding paradox in neural machine translation: the sequence with the highest probability is not equivalent to the sequence with the best translation quality, and degenerate phenomena such as empty or excessively short sequences may even occur.
1
Sources
50%
Confidence
Long-term
Relevance
8/22/2026
First Seen
Sources
Related Entities
Related Claims
Unverified大语言模型本质上是概率性的，相同的输入可能产生不同的输出，这与传统软件的确定性行为截然不同65% similarUnverifiedText generated by large language models tends to have low perplexity, with word choices that are overly consistent with probabilistic expectations, lacking the unexpectedness of human writing.64% similarUnverifiedTraditional n-gram models and neural language models trained on a single language assign extremely low probability scores to cross-lingual word sequences, causing the decoder to tend to forcibly map foreign-language words to phonetically similar words in the target language.64% similarUnverifiedLarge language models autoregressively predict the next highest-probability token, causing AI-generated text to have low perplexity.64% similarUnverifiedLarge language models are fundamentally probabilistic and cannot guarantee 100% output correctness.63% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/786974API
curl https://kongchang.com/api/v1/knowledge/claims/786974MCP
get_claim(id=786974)