Logprob

Definitions

1. The logarithm of a probability assigned by a model to a token or sequence, commonly used for numerical stability in scoring and decoding. Logprobs are useful for analyzing confidence, ranking alternatives, and debugging generation behavior.

   In plain English: A log-scaled probability score used to analyze how likely a model thinks some output is.

   Example: "They inspected token logprobs to understand why the model kept favoring one brand name over a nearly identical competitor."

Related Terms

• CPT
• Hidden Markov Model
• Length Penalty
• Likelihood
• Model Temperature
• Penalty
• Repetition Penalty
• Self-Consistency
• Stop Sequence
• Token Probability