Skip to content

bmel:softmax

Category: LLM & AI Observability · Returns: bmel:array<number>

bmel:softmax(logits: bmel:array<number>)

Description

Applies the softmax function to an array of numeric logits, returning a probability distribution where all values are in (0, 1) and sum to exactly 1.0. Formula: softmax(x_i) = exp(x_i) / sum(exp(x_j)). Used in Transformer attention scoring and for converting raw model logits into interpretable confidence scores.

Arguments

Parameter Type Required Description
logits bmel:array<number> Array of raw logit values (unnormalized log-probabilities) to convert into a probability distribution.

Example

bmel:softmax({model:Response Payload}.$.logits)

Back to BMEL Reference