Skip to main content
Back to challenges
Challenge

OpenAI Agents Litigation Engine: Nuclear Verdict Risk Predictor

Surging litigation costs and social inflation create unexpected catastrophic losses for casualty insurers across civil jurisdictions. Build an OpenAI Agents SDK system to extract legal case factors, predict nuclear verdict probability, and maintain a Brier score below 0.12 on liability claims.

Machine LearningHosted by Vera
Sign in free to start
Challenge brief

What you are building

The core problem, expected build, and operating context for this challenge.

Build an agent using the OpenAI Agents SDK to analyze casualty claim dockets and estimate high-severity litigation probability.

Delivery guide

How work is evaluated

Evaluation

Evaluates nuclear verdict probability estimation calibration and accuracy.

Datasets

Challenge datasets

Public datasets, plus the private datasets you have access to.

Evaluation rubric

How submissions are scored

These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.

Dimensions
2 scoring checks
Binary
2 pass or fail dimensions
Ordinal
0 scaled dimensions
Dimension 1

brier_score_threshold

Ensures prediction Brier score is below 0.12.

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Dimension 2

brier_score

Brier score measuring probability calibration accuracy. • target: 0.1 • range: 0-1

Binary check

This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.

Learning goals

What you should walk away with

  • Implement autonomous agent reasoning loop using OpenAI Agents SDK.

  • Extract litigation risk factors (venue jurisdiction, plaintiff counsel profile, injury severity).

  • Predict probability of a nuclear verdict (> $10M liability award).

  • Achieve a calibrated Brier score under 0.12 across test cases.

Resources and assets

Reference links and supporting material

Dataset notes

Dataset of 300 commercial liability civil court cases with actual final jury award values.

Challenge setup

Evaluates nuclear verdict probability estimation calibration and accuracy.

Challenge input

JSON containing case_id, venue_county, injury_type, defense_reserve_usd

OpenAI Agents SDK

Official OpenAI SDK for autonomous multi-tool agent orchestration.

Evaluated output

JSON containing nuclear_verdict_probability, key_drivers, recommended_reserve_usd

Evaluation checks
  • Ensures prediction Brier score is below 0.12.
  • Brier score measuring probability calibration accuracy. • target: 0.1 • range: 0-1
Configured evaluation evidence
  • Benchmark: Legal Risk Calibration Benchmark
  • Brier Score target: 0.1
  • 1 public reference case
Configured runtime
  • Python execution harness
  • Python sandbox (unavailable on Versalist)

Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.

View technical recipe

Configured tools

Action Space
  • OpenAI · Required
  • Llama 3.3 70B · Optional
  • Azure OpenAI Service · Optional

Evaluation contract

  • brier_score_threshold · Weight 1
  • brier_score · Weight 1

Metrics

  • brier_score · Target 0.1

Benchmarks

  • Legal Risk Calibration Benchmark
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing case_id, venue_county, injury_type, defense_reserve_usd

    Output: JSON containing nuclear_verdict_probability, key_drivers, recommended_reserve_usd

    Execution harness: Python

Evaluation tools

  • OpenAI Agents SDK · Agent Framework

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about OpenAI Agents Litigation Engine: Nuclear Verdict Risk Predictor