Original prompt
Draw an SVG of a person pushing, not pulling, a wheelbarrow.
推独轮车的人
4 runs · 07.A.01 · SVG
PROMPT/Phase1_TEST_PROMPTS_BILINGUAL.md · The original prompt is kept verbatim
Draw an SVG of a person pushing, not pulling, a wheelbarrow.
绘制一幅一个人正在推、而不是拉独轮车的 SVG。
The deliverable should be a standalone SVG that clearly places the person behind the wheelbarrow handles, facing the direction of travel and gripping forward to communicate pushing. The wheelbarrow should be recognizable through a single front wheel, tray, supports, and two handles. A leaning torso, walking stride, or force relationship may reinforce forward motion. The composition must avoid pulling cues such as placing the person in front, facing backward, or dragging the handles behind them.
All three results use the same width and the same preview ratio; no model is marked as best.
Cross-model comparison must note differences in harness, budget and isolation.
| Metric | DeepSeek-V4.1-Flashrun-deepseek-v4-1-flash-exp-0910-task-07-r1 | Muse Spark 1.3run-muse-spark-1-3-task-07-r1 | GPT-5.6 Solrun-gpt-5-6-sol-task-07-r1 | K3run-k3-task-07-r1 |
|---|---|---|---|---|
| Runtime | 0:48 | 0:40 | 3:09 | 4:19 |
| API calls | 3 | 7 | 12 | 5 |
| Tool calls | 2 | 7 | 11 | 4 |
| Failures / retries | 0 | 0 | 1 | 0 |
| Token throughput | 248.9K | 96.2K | 363.4K | 163.5K |
| AI review | 95/100 | 86/100 | 92/100 | 95/100 |
Two runs of the same task version can be compared side by side.
Type to search; ↑ ↓ to choose, Enter to open
No matching archive records.