Eval Results

1 个变体

An AI eval results card that compares models across test cases, streams a run in cell by cell, and flags regressions.

Model comparison

eval-results.tsx

Pass rates roll up and split bars grow as each result lands. Filter to failing cases or regressions, and open any case to read every model's output with the grader's note.

Support replies

Run 42 · 8 cases

Atlas 2

Large

88%pass rate

+7 vs last run

Kite

Small

50%pass rate

-6 vs last run

Lumen

Fine-tuned

38%pass rate

-12 vs last run

Test case	Atlas 2	Kite	Lumen
Billing	.94	.61	.88
ToneRegression	.91	.22	.58
Product	.97	.90	.86
SafetyRegression	.96	.89	.10
BillingRegression	.64	.84	.55
Tone	.93	.52	.87
Product	.90	.18	.60
Billing	.92	.85	.30

TSXcomponents/previews/premium/eval-results-preview.tsx

"use client";

import { EvalResults } from "@/components/premium/eval-results";

export function EvalResultsPreview() {
  return (
    <div className="w-full px-4 py-10">
      <EvalResults />
    </div>
  );
}

安装

$ bunx --bun shadcn add @beui-pro/

Eval Results

Model comparison

源代码已锁定

安装