mirror of
https://github.com/EvoMap/evolver.git
synced 2026-09-18 21:47:53 +08:00
evolver-publish
25 lines
1.3 KiB
JavaScript
25 lines
1.3 KiB
JavaScript
/** value = severity × reach × strategicFit × (1+novelty)/(1+costEst). 命名因子, 禁黑盒 (军杰§5.3). */
|
||
export function computeValue(f) {
|
||
const score = (f.severity * f.reach * f.strategicFit * (1 + f.novelty)) / (1 + f.costEst);
|
||
return { score, factors: { ...f } };
|
||
}
|
||
/** occurrences that earn full reach — a FIXED scale, deliberately NOT normalized to the biggest bucket. */
|
||
export const REACH_FULL_AT = 30;
|
||
const UNCLASSIFIED_FIT = 0.35;
|
||
const CLASSIFIED_FIT = 0.9;
|
||
const clamp01 = (n) => Math.max(0, Math.min(1, n));
|
||
/**
|
||
* Derive value factors from a problem's observation stats — the tuning learned from observing real agent logs
|
||
* (evolver-v2-observation): a vague high-volume catch-all should not out-trigger specific, actionable problems.
|
||
* So `reach` is on a FIXED occurrence scale (not normalized to the largest bucket, which crushes everything
|
||
* specific), and `strategicFit` is reduced for unclassified problems. Feed the result to {@link computeValue}.
|
||
*/
|
||
export function deriveValueFactors(stats) {
|
||
return {
|
||
severity: clamp01(stats.severity ?? 0.5),
|
||
reach: clamp01(Math.max(0, stats.occurrences) / REACH_FULL_AT),
|
||
strategicFit: stats.classified === false ? UNCLASSIFIED_FIT : CLASSIFIED_FIT,
|
||
novelty: stats.novelty ?? 0,
|
||
costEst: stats.costEst ?? 0.2,
|
||
};
|
||
} |