Legal Accuracy
25 ptsIs the substantive legal answer correct?
Proprietary Evaluation Methodology
LEX-EVAL™
A human-led Legal AI Evaluation Framework designed to distinguish fluent language from actual legal reliability.
“High linguistic quality is not evidence of legal reliability.”
Decision Layer
90–100
Expert Grade
80–89
Strong
70–79
Conditional
50–69
High Risk
0–49
Unreliable
CLF
Overrides score
Reliability outcome: PASS · PASS WITH REVIEW · FAIL-CRITICAL.
100-Point Architecture
Seven dimensions. One auditable evaluation record.
Authority & Citation Integrity
15 ptsAre authorities real, relevant, current, and properly represented?
Issue Identification
10 ptsDoes the output identify the legal issues that actually matter?
Legal Reasoning
20 ptsDoes the analysis connect facts, rules, authorities, and conclusions coherently?
Jurisdictional Alignment
10 ptsDoes the answer remain within the correct legal system and applicable framework?
Hallucination & Fabrication Control
10 ptsDoes it avoid invented authorities, facts, rules, and unsupported certainty?
Professional Usefulness
10 ptsCan a professional reviewer use the output efficiently and safely?
Critical Legal Failures
A numerical score cannot neutralize a Critical Legal Failure. LEX-EVAL™ separates aggregate quality from failures that can make professional reliance unsafe.
CLF-01Fabricated Authority
CLF-02Material Misstatement of Law
CLF-03Wrong Jurisdiction
CLF-04Invalid or Obsolete Authority
CLF-05Material Fact Distortion
CLF-06Unsafe Professional Reliance
Evaluation Deliverables
From score to corrective intelligence.
Weighted scorecard and reliability outcome
Authority and citation verification
Error taxonomy and severity classification
Critical Legal Failure assessment
Corrected or gold-standard answer
Aggregate benchmark analytics