Toor

Appearance

@gettoor/core v0.1.0

@gettoor/core / ModelParameters

Interface: ModelParameters ​

Model parameters for the LLM-as-a-judge.

Extended by ​

Properties ​

frequencyPenalty? ​

optional frequencyPenalty?: number

Penalizes tokens based on how frequently they have already appeared in the generated text.

Higher values reduce repetition of the same words and phrases, resulting in more varied output.

maxOutputTokens? ​

optional maxOutputTokens?: number

Maximum number of tokens the model may generate in its response.

Lower values limit response length and cost, while higher values allow longer and more detailed outputs.

presencePenalty? ​

optional presencePenalty?: number

Penalizes tokens that have already appeared in the generated text.

Higher values encourage the model to introduce new topics and vocabulary instead of repeating previously used tokens.

temperature? ​

optional temperature?: number

Controls sampling randomness.

Lower values make outputs more deterministic and focused, while higher values increase creativity and variation.

topK? ​

optional topK?: number

Controls top-k sampling.

The model samples only from the topK most likely next tokens. Lower values produce more predictable outputs, while higher values allow for more varied and creative responses.

topP? ​

optional topP?: number

Controls nucleus sampling.

The model considers the smallest set of tokens whose cumulative probability is at least topP and samples from that set. Lower values make outputs more focused and deterministic, while higher values increase diversity.