HALOWERKAialignwerk
goal drift
Aligns caller-labeled dimensions, computes cosine similarity and a normalized L1 difference, and flags drift against a supplied cosine threshold. It does not infer goals from behavior, decide which goal is correct or detect deception; results depend entirely on the supplied vector representation.
POST https://aialign.halowerk.com/v1/goal-driftEingaben
| Feld | Typ | Pflicht | Beschreibung |
|---|---|---|---|
| dimensions | array | Ja | |
| minimum_cosine_similarity | number | Ja |
Vollständiges Eingabeschema anzeigen
{
"type": "object",
"required": [
"dimensions",
"minimum_cosine_similarity"
],
"additionalProperties": false,
"properties": {
"dimensions": {
"type": "array",
"minItems": 1,
"maxItems": 10000,
"items": {
"type": "object",
"required": [
"name",
"initial_weight",
"current_weight"
],
"additionalProperties": false,
"properties": {
"name": {
"type": "string",
"minLength": 1,
"maxLength": 128
},
"initial_weight": {
"type": "number",
"minimum": -1000000000000,
"maximum": 1000000000000
},
"current_weight": {
"type": "number",
"minimum": -1000000000000,
"maximum": 1000000000000
}
}
}
},
"minimum_cosine_similarity": {
"type": "number",
"minimum": -1,
"maximum": 1
}
}
}Veröffentlichtes Eingabebeispiel
{
"dimensions": [
{
"name": "user_safety",
"initial_weight": 0.7,
"current_weight": 0.62
},
{
"name": "task_completion",
"initial_weight": 0.2,
"current_weight": 0.31
},
{
"name": "resource_efficiency",
"initial_weight": 0.1,
"current_weight": 0.07
}
],
"minimum_cosine_similarity": 0.97
}Ergebnis & Dokumentation
Veröffentlichtes Ergebnisschema anzeigen
{
"$ref": "https://aialign.halowerk.com/v1/goal-drift/schema#/outputSchema"
}Veröffentlichtes Ergebnisbeispiel öffnen
Maschinenlesbare Dokumentation dieses Hubs
Quelle dieser Seite: der veröffentlichte HALOWERK-Katalog. Stand: 2026-11-01, 15:19 UTC.
Weitere Endpunkte dieses Hubs
- Aialignwerk0,005 USDC
agent collusion
POSThttps://aialign.halowerk.com/v1/agent-collusionRequires the same agents in each supplied round, computes each pair’s observed action agreement, expected agreement from marginal action frequencies and chance-corrected kappa, then flags high positive agreement against a supplied threshold
- Aialignwerk0,004 USDC
bias audit
POSThttps://aialign.halowerk.com/v1/bias-auditCalculates each group positive-outcome rate, compares it with an explicit or automatically selected reference group, and reports rate differences and selection-rate ratios. Aggregate disparity metrics do not establish discrimination, fairne
- Aialignwerk0,003 USDC
capability eval
POSThttps://aialign.halowerk.com/v1/capability-evalComputes weighted pass rates and weighted scores from caller-labeled test outcomes, with deterministic category summaries. It does not run tests, validate labels, measure untested capabilities or establish deployment safety; the result is o