Extracts token log probabilities from assistant replies within an LLMMessage
object.
Each row represents a token with its log probability and top alternative tokens.
Details
An empty tibble is output if no logprobs were requested. Currently only works with openai_chat()
Columns include:
reply_index
: The index of the assistant reply in the message history.token
: The generated token.logprob
: The log probability of the generated token.bytes
: The byte-level encoding of the token.top_logprobs
: A list column containing the top alternative tokens with their log probabilities.