use-ml-evaluation

Evaluation run and validation metadata primitives for RustUse.

Experimental

use-ml-evaluation is experimental while use-ml remains below 0.3.0.

Example

use use_ml_evaluation::{MlEvaluationKind, MlEvaluationRunId, MlThreshold};

let run_id = MlEvaluationRunId::new("eval-001")?;
let threshold = MlThreshold::new(0.5)?;
let kind: MlEvaluationKind = "cross-validation".parse()?;

assert_eq!(run_id.as_str(), "eval-001");
assert_eq!(threshold.value(), 0.5);
assert_eq!(kind, MlEvaluationKind::CrossValidation);
# Ok::<(), use_ml_evaluation::MlEvaluationError>(())

Scope

• Evaluation run IDs, kinds, validation strategies, targets, statuses, slices, and benchmarks.
• Threshold metadata and confusion-matrix shape metadata.
• Generic ML evaluation metadata only.

Non-goals

• Computing evaluation metrics beyond trivial metadata validation.
• LLM-as-judge, prompt evaluation, conversation evaluation, hallucination checks, safety/guardrail evaluation, or retrieval-groundedness evaluation.

License

Licensed under either Apache-2.0 or MIT.