Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →A股GAN模拟/对抗生成数据增强。当用户说"GAN"、"对抗生成"、"数据增强"、"生成对抗"、"合成数据"、"模拟数据"、"GAN模拟"时触发。基于 cn-stock-data 获取数据，使用GAN生成合成金融数据。支持 formal/brief 两种输出风格。
.claude/skills/aifinlab-a-share-gan-simulation/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-07 | ✗→✓ | ▲ Improved | 18% | 0% |
| case-11 | ✗→✓ | ▲ Improved | -62% | 0% |
| case-03 | ✓→✓ | ＝ Same ✓ | 10% | 0% |
| case-04 | ✓→✓ | ＝ Same ✓ | 33% | 0% |
| case-05 | ✓→✓ | ＝ Same ✓ | -4% | 0% |
通过 cn-stock-data skill 获取数据：
# GAN数据增强报告
## 一、生成配置
| 参数 | 设置 |
|------|------|
## 二、数据质量
[统计对比、时序检验]
## 三、增强效果
[模型表现提升]
## 四、应用建议## GAN增强速览
- TimeGAN生成5000条合成序列
- 统计特性匹配度 92%
- 数据增强后模型IC提升 +8%
- 极端场景覆盖率提升至95%参考 references/gan-simulation-guide.md 获取详细方法论与 A股实证研究。
| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→fail | 47,569 | 37,573 | -21% | 1 | 1 | 0% | 6,954 | 5,359 | -23% | 0 | 0 | — |
case-02 | fail→fail | 25,381 | 11,944 | -53% | 1 | 1 | 0% | 3,799 | 1,260 | -67% | 0 | 0 | — |
case-03 | pass→pass | 35,629 | 37,790 | +6% | 1 | 1 | 0% | 5,770 | 6,332 | +10% | 0 | 0 | — |
case-04 | pass→pass | 28,929 | 33,050 | +14% | 1 | 1 | 0% | 5,164 | 6,864 | +33% | 0 | 0 | — |
case-05 | pass→pass | 28,331 | 27,330 | -4% | 1 | 1 | 0% | 5,471 | 5,227 | -4% | 0 | 0 | — |
case-06 | pass→pass | 15,889 | 13,668 | -14% | 1 | 1 | 0% | 3,053 | 3,246 | +6% | 0 | 0 | — |
case-07 | fail→pass | 26,593 | 26,997 | +2% | 1 | 1 | 0% | 3,532 | 4,160 | +18% | 0 | 0 | — |
case-08 | pass→pass | 20,825 | 23,490 | +13% | 1 | 1 | 0% | 3,123 | 3,628 | +16% | 0 | 0 | — |
case-09 | pass→pass | 12,486 | 16,474 | +32% | 1 | 1 | 0% | 2,064 | 3,092 | +50% | 0 | 0 | — |
case-10 | pass→pass | 18,984 | 26,144 | +38% | 1 | 1 | 0% | 2,865 | 4,447 | +55% | 0 | 0 | — |
case-11 | fail→pass | 15,203 | 2,709 | -82% | 1 | 1 | 0% | 2,371 | 912 | -62% | 0 | 0 | — |
case-12 | pass→pass | 20,233 | 22,671 | +12% | 1 | 1 | 0% | 3,254 | 4,199 | +29% | 0 | 0 | — |
case-13 | pass→pass | 19,763 | 19,777 | +0% | 1 | 1 | 0% | 3,148 | 3,814 | +21% | 0 | 0 | — |
case-14 | pass→pass | 23,708 | 25,501 | +8% | 1 | 1 | 0% | 3,554 | 3,764 | +6% | 0 | 0 | — |
case-15 | pass→pass | 21,564 | 24,570 | +14% | 1 | 1 | 0% | 3,063 | 4,119 | +34% | 0 | 0 | — |
case-16 | pass→pass | 24,626 | 28,577 | +16% | 1 | 1 | 0% | 3,220 | 4,377 | +36% | 0 | 0 | — |
case-17 | pass→pass | 20,306 | 21,610 | +6% | 1 | 1 | 0% | 2,928 | 3,267 | +12% | 0 | 0 | — |
case-18 | pass→pass | 21,671 | 19,808 | -9% | 1 | 1 | 0% | 2,950 | 3,563 | +21% | 0 | 0 | — |
case-19 | pass→pass | 9,961 | 4,453 | -55% | 1 | 1 | 0% | 1,607 | 1,140 | -29% | 0 | 0 | — |
case-20 | pass→pass | 19,974 | 22,074 | +11% | 1 | 1 | 0% | 2,996 | 3,351 | +12% | 0 | 0 | — |
case-21 | pass→pass | 17,541 | 21,043 | +20% | 1 | 1 | 0% | 2,612 | 3,570 | +37% | 0 | 0 | — |
case-22 | pass→pass | 19,875 | 15,202 | -24% | 1 | 1 | 0% | 2,611 | 2,874 | +10% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 22 cases were attempted, and 21 counted toward the lift figure. The other 1 produced results that are not comparable between the two arms, so they are excluded from the headline rather than averaged into it. The headline lift of +9 percentage points is the difference between those two pass rates over the 21 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
| Model | Method | Date | Lift |
|---|---|---|---|
| gemini-3.6-flash | verified | 8/28/2026 | +12% |
Other measured skills in the registry, with their headline benchmark lift.