Will decoding always preserve the original capitalization?
Only if the tokenizer preserved that information during encoding and the decoder retains it. A tokenizer that lowercases input cannot automatically recover the original capitalization.
Explore This Question
Search for additional explanations using these tools:
More FAQs in How are tokens converted back into human-readable text?
No. Generation selects a sequence of token IDs using the model and decoding strategy. Tokenizer decoding converts that selected sequence into a readable string.
Some tokens represent incomplete byte sequences or depend on surrounding pieces for spacing. Decoding the sequence together helps reconstruct characters and boundaries correctly.
Still have questions?
Browse all our FAQs or reach out to our support team