Back to evaluations
Public evaluation
nuclear_verdict_probability
Evaluates probability score calibration for high-severity legal nuclear verdicts.
Evaluation type
task based
Challenge
Legal Nuclear Verdict Predictor using LangChain and LlamaIndex
Difficulty
Advanced
Rigor
Unspecified
Evaluation overview
How the linked challenge is judged: tasks, benchmarks, and criteria count.
Tasks
1
Benchmarks
0
Criteria
0
Task templates
Inputs and expected outputs.
Task 1
nuclear_verdict_probability
Predicts nuclear verdict likelihood for a casualty claim file.
Input format
JSON containing claim_summary, jurisdiction, plaintiff_demands_m, and injury_severity_code
Output format
JSON with nuclear_verdict_prob (float 0-1), primary_risk_jurisdiction (string), and expected_severity_tier (string)