Logprobs evaluate token confidence by taking the mathematical exponent of each candidate token's log-probability to calculate a normalized percentage score, identifying both the model's certainty and high-probability alternative branching paths.
flowchart LR
A["Prompt Input<br/>'Brand is known for...'"] --> B["Token: 'shoes'<br/>Logprob: -0.105 (90.0%)"]
A --> C["Token: 'boots'<br/>Logprob: -2.302 (10.0%)"]
B --> D[High Confidence Path]
C --> E[Alternative Branch]