Aider Polyglot
Harbor nativeAider Polyglot: 6种语言diff编辑
Verified71%Coding
- Samples
- 225
- Released
- 2024
- Scores
- 61
WebUIProof
WebUIProof: Benchmarking WebUI Code Generators with UI-Agent Execution Harness
An execution-oriented benchmark with 219 tasks and 3k executable interaction tests measuring functional correctness of WebUI code generation via a UI-agent harness.
How to run
ltb · webuiproof
$ npm i -g lettokenburn-cli$ ltb bench info webuiproof
External harness. No Harbor adapter has been generated for this benchmark yet. `ltb bench info` prints the known links.
Runs happen on your machine with your own subscription; only the score summary is pushed. CLI docs
Scores
10 sourced entries| Harness | Source | Confidence | Link | |||
|---|---|---|---|---|---|---|
| Claude Sonnet 4 (2025-05-14) | 46.1 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Gemini 2.5 Pro | 42.9 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Qwen 3 Coder 480B | 42.0 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| DeepSeek R1 | 41.9 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| GPT-4.1 | 41.4 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Llama 4 Maverick | 40.8 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Kimi K2 Thinking Turbo | 39.3 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Llama 3 405B | 36.2 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| Qwen2.5 14B Instruct | 26.1 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
| MIMO 7B | 23.8 | N/A | Independent | medium | Oct 2, 2026 | arxiv.org |
Every number links to where it was published. Low-confidence rows are muted and need per-cell verification.
Reproduced by the community
All runsLoading community runs…
About
WebUIProof contains 219 tasks (149 general, 70 3D) with structured specifications and 3,000 executable interaction tests. It evaluates WebUI code generators by running tests in a headless browser using an iterative plan-act-observe loop.
Scoring relies on the success rate of UI-agent executed assertions checking DOM changes and behavior. Evaluation of eight commercial LLMs reveals frequent failures on interaction requirements despite successful rendering, especially for 3D simulations.