Hybrid IRAC Scoring

Blending deterministic chain consistency with expert-calibrated synthesis rubrics (MEE).

The Composite Method

Evaluating legal synthesis is notoriously difficult due to the nuance of reasoning. LegalChain solves this by using a Composite Scoring Architecture that balances three distinct signals, minimizing the risks of "LLM-as-a-Judge" circularity.

Signal 01
10%
Structure
Signal 02
40%
Consistency
Signal 03
50%
Reasoning

Structure (Presence of IRAC) + Consistency (Fact Usage) + Quality (MEE Rubric)

MEE-Based Reasoning Rubric

The 50% "Reasoning Quality" signal is assessed using a rubric derived from the **Multistate Essay Examination (MEE)** professional standards. Our LLM-Judge is calibrated to evaluate:

I

Issue Identification

Does the model capture the precise legal conflict identified in S1?

R

Rule Statement

Is the legal principle extracted from the cited case accurate and complete?

A

Application / Analysis

The "Reasoning Bridge": Does the model logically connect facts to the rule?

C

Conclusion

Is the final outcome grounded in the preceding analysis?

Deterministic Anchoring

The 40% **Chain Consistency** component is deterministic. We check if the model used the specific metadata and holding codes extracted in steps S1 and S4. If a model writes a "correct" answer but ignores its own previous findings, it is penalized for lack of internal coherence—a hallmark of stochastic mimicry rather than logical reasoning.