Returns the eight dimensions every RAIL evaluation scores, each with a description and the score bands a result falls into. Each dimension also carries the weight and threshold configured for your application, so you can see exactly how scores are aggregated and what counts as a pass.
Request
curl https://api.responsibleailabs.ai/railscore/v1/dimensions \
-H "Authorization: Bearer YOUR_RAIL_API_KEY"
Response
{
"dimensions": [
{ "key": "fairness", "name": "Fairness", "description": "Bias, stereotyping, and equitable treatment across groups.", "weight": 12.5, "threshold": null },
{ "key": "safety", "name": "Safety", "description": "Harmful, dangerous, or otherwise unsafe content.", "weight": 20, "threshold": 8 },
{ "key": "reliability", "name": "Reliability", "description": "Factual accuracy and freedom from hallucination.", "weight": 12.5, "threshold": null }
],
"score_bands": [
{ "band": "Excellent", "min": 9.0 },
{ "band": "Good", "min": 7.0 },
{ "band": "Needs improvement", "min": 5.0 },
{ "band": "Poor", "min": 3.0 },
{ "band": "Critical", "min": 0.0 }
]
}
The eight RAIL dimensions. Each entry has:
key — the identifier used in evaluation dimension_scores (e.g. user_impact).name — the display name used in policy weights and thresholds (e.g. User Impact).description — what the dimension measures.weight — how much this dimension contributes to the overall score for your application.threshold — the per-dimension minimum for your application, or null if no per-dimension gate is set.
The qualitative bands a score (0–10) maps to, from Excellent down to Critical.
