transfer-form
0 design-system2 qualityIncompletenvidia/nemotron-3.5-lightning-free · genui-spec · all runs of transfer-form · Nov 10, 16:08
Run
- Run
- e9d384e5
- Cost
- $0.0000 (gateway $0.0000)
- Duration
- 29.4 s
- Steps
- 3 of 20 allowed, ended with tool-calls
- Tokens
- 9,610 in (12,576 cached), 6,282 out
- Page
- Rendered
Batch
Runs from different design-system versions should not be compared.
- Label
- behaviour-v3
- Commit
- 802950f
- Design system
The page
A screenshot in each theme, as the benchmark took it.
page.tsx
As the agent left it; marked lines broke a rule.
import { GeneratedView } from "@vektos/genui";Violations
2 counted, under 2 rules. “Not counted” marks a compiler error that restates a design-system violation, “Ignored” one the agent silenced with a comment.
- page.tsx:1:1An amount above the available balance shows an error on the amount field. Not done: the field "Amount" is not marked invalid (aria-invalid, or the browser's own check).
- page.tsx:1:1A valid transfer shows a confirmation with the beneficiary. Not done: "Studio Legale Ferri" is not on screen.
Steps
One row per model call, with the tools it asked for. An agent that reads the guidelines before writing shows it here; what a checker or submit_view answered is under its call.
| Step | Tools | Tokens in | Tokens out | Cost |
|---|---|---|---|---|
| 0 | submit_view submit_view: These parts of your view could not be shown. Send the whole view again with them fixed: - $.children[1].type: Put each SelectField inside a Form. | 1,913 | 4,430 | $0.0000 |
| 1 | submit_view submit_view: These parts of your view could not be shown. Send the whole view again with them fixed: - $.children[2].children[0].props.value: `value` must be one of the `options`. | 6,387 | 1,142 |