Skip to main content
Back to challenges
Challenge

Predict Nuclear Verdict Litigation Risk with Claude Agents SDK and Novita AI

Surging legal liabilities and nuclear verdicts present extreme financial risk to commercial general casualty underwriters. Build an automated legal severity scoring pipeline using Claude Agents SDK hosted on Novita AI runtime. Achieve a 0.85+ ROC-AUC score in identifying high-risk litigated bodily injury 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.

Construct an agent system using Claude Agents SDK deployed via Novita AI runtime to analyze legal filings and predict potential nuclear verdict escalation.

Delivery guide

How work is evaluated

Evaluation

Evaluates predicted nuclear verdict likelihood scores against historical trial outcome data.

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

valid_probability_output

Probability score lies between 0.0 and 1.0.

Binary check

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

Dimension 2

roc_auc

Area under ROC curve for binary nuclear verdict prediction (>10M threshold) • target: 0.85 • 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 multi-stage extended reasoning pipelines using Claude Agents SDK.

  • Deploy models via Novita AI serverless model execution environment.

  • Extract litigation markers from legal complaint documents.

  • Calculate calibrated nuclear verdict probability scores.

Resources and assets

Reference links and supporting material

Dataset notes

1,000 anonymized court litigation summaries with ground-truth judgment amounts.

Challenge setup

Evaluates predicted nuclear verdict likelihood scores against historical trial outcome data.

Challenge input

JSON containing case docket summary, venue, plaintiff attorney, injury severity

Claude Agents SDK

Provides native support for tool orchestration and complex reasoning.

Novita AI

Offers rapid serverless model execution endpoints.

Evaluated output

JSON with nuclear_verdict_prob, key_risk_drivers, and recommended_reserve_usd

Evaluation checks
  • Probability score lies between 0.0 and 1.0.
  • Area under ROC curve for binary nuclear verdict prediction (>10M threshold) • target: 0.85 • range: 0-1
Configured evaluation evidence
  • Benchmark: LegalRiskBench
  • Roc Auc target: 0.85
  • 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
  • Novita AI · Required
  • Phi-3 · Optional
  • Writer · Optional

Evaluation contract

  • valid_probability_output · Weight 1
  • roc_auc · Weight 1

Metrics

  • roc_auc · Target 0.85

Benchmarks

  • LegalRiskBench
  • general

Reference cases

  • 1 public reference case

Workflow tasks

  1. Task 1

    Input: JSON containing case docket summary, venue, plaintiff attorney, injury severity

    Output: JSON with nuclear_verdict_prob, key_risk_drivers, and recommended_reserve_usd

    Execution harness: Python

Evaluation tools

  • Claude Agents SDK · Agent Framework
  • Novita AI · Inference Host

Recipe freeze marker

No recipe freeze time is recorded.

Frequently Asked Questions about Predict Nuclear Verdict Litigation Risk with Claude Agents SDK and Novita AI