JEV USE CASE
Citation Verification.
Compare a claim, its cited passage and source context to classify whether the citation supports the claim.
THE DECISION CONTRACT
STATE
Claim: refunds rose 18%. Source: refunds rose from 10% to 11.8%.
QUESTION
What is the relationship between the source and claim?
CHOICES
supports partially_supports contradicts insufficient
TYPESCRIPT EXAMPLE
Ask for one bounded answer
This Vercel AI SDK-shaped example is illustrative. Check the current provider documentation before shipping.
import { experimental_evaluate as evaluate } from 'ai';
const result = await evaluate({
model: 'typesafe-ai/jev',
state: "Claim: refunds rose 18%. Source: refunds rose from 10% to 11.8%.",
questions: {
decision: {
type: 'choice',
instructions: "What is the relationship between the source and claim?",
options: ["supports","partially_supports","contradicts","insufficient"]
}
}
});
// Validate permissions and thresholds before acting.
console.log(result.answers.decision);SAFE IMPLEMENTATION
Keep the boundary clear
Guardrail
Run exact quote and source checks in code first. Route ambiguous semantic judgments to review.
- Log the input, answer distribution, chosen threshold and resulting route.
- Test on labelled examples from your own production distribution.
- Send low-confidence, high-impact or out-of-distribution cases to a safer fallback.
- Re-evaluate thresholds when the model or workflow changes.
PRIMARY SOURCES