兩項目評估報告
- EvalRun: #292(suite #147
auto-fd7bdd03-r1-053959809)
- Target: #2 production-canary
- Cases: 30(N=1, attempts=30/30)
- Generated: 2026-05-16T14:15:59+08:00
本報告含兩個獨立評估項目。項目一 評估 bot 進入「知識與產品查詢」fallback
後的撈取 + 回答能力;項目二 評估 bot 跨全 enabled scenarios 的 routing →
tool → answer 三段 funnel 健康度。
項目一: 知識庫精準度
評估範圍 (prerequisite: 進對 scenario)
- 進對 scenario 的 attempts: 10 / 30 (33.3%)
- 只對這些 attempts 算 retrieval + answer
- 沒進對 scenario 的 attempts → 排除(routing 失敗不該污染 KB 訊號)
Per-scenario 評估
| Scenario |
qualifying / total attempts |
retrieval_relevance |
answer_correctness |
| 知識與產品查詢 |
10 / 10 |
⚪ 53.3% |
✅ 100.0% |
待修方向(worst-3)
- 知識與產品查詢 × retrieval_relevance = 53.3% — retrieval 略低 — 看 LLM relevance vs topic_match 哪邊弱
- 知識與產品查詢 × answer_correctness = 100.0% — 回答內容有問題 — 細看 hallucination_rate vs answer_quality 哪邊低
項目二: 情境調用與完成
整體 funnel(全 scenarios 加總)
| Stage |
Pass count |
% of total |
| Total attempts |
20 |
100.0% |
| Step 1: scenario_routing |
4 |
20.0% |
| Step 2: + tool_calling |
4 |
20.0% |
| Step 3: + answer |
1 |
5.0% |
Per-scenario funnel
| Scenario (n_attempts) |
Step 1 (routing) |
Step 2 (tool) |
Step 3 (answer) |
end-to-end |
| early_return (9) |
❌ 0/9 (0.0%) |
❌ 0/9 (0.0%) |
❌ 0/9 (0.0%) |
❌ 0/9 (0.0%) |
| 女性更年期保養 (1) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
| 媽媽禮活動字串 (1) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
| 小產 (1) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
| 提供產品名稱 (1) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
| 男性一般保養 (1) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
| 男性備孕 (1) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
| 知識與產品查詢 (1) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
| 詢問出貨時間 (1) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
| 詢問吃產品會胖嗎? (1) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
| 詢問退貨 (2) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
Drop-off 最大的 5 個 scenario
- early_return drop at step1 (-100.0pp)
- 女性更年期保養 drop at step3 (-100.0pp)
- 媽媽禮活動字串 drop at step3 (-100.0pp)
- 小產 drop at step1 (-100.0pp)
- 提供產品名稱 drop at step1 (-100.0pp)
Audit
Reproduce combined report: bin/rails runner "puts Eval::EvaluationReport.call(run: EvalRun.find(292))"
Or fetch each item separately:
bin/rails runner "puts Eval::KbAccuracyReport.call(run: EvalRun.find(292))"
bin/rails runner "puts Eval::ScenarioFunnelReport.call(run: EvalRun.find(292))"
Per-Scenario × Per-Dim — Run #292
Suite: 亞尼活力_測試用 (bulk R1) · scenarios: 11 · dims: 5 · populated cells: 53/55
| Scenario |
Scenario |
Tool |
Retrieval |
Faith |
AnsQ |
| early_return |
0.0% [0.0–0.0] (n=9) |
0.0% (n=1) |
87.5% [62.5–100.0] (n=8) |
100.0% [100.0–100.0] (n=9) |
85.6% [67.0–99.3] (n=9) |
| 女性更年期保養 |
100.0% (n=1) |
0.0% (n=1) |
0.0% (n=1) |
100.0% (n=1) |
43.3% (n=1) |
| 媽媽禮活動字串 |
100.0% (n=1) |
— |
0.0% (n=1) |
0.0% (n=1) |
46.7% (n=1) |
| 小產 |
100.0% (n=1) |
0.0% (n=1) |
0.0% (n=1) |
0.0% (n=1) |
83.3% (n=1) |
| 提供產品名稱 |
100.0% (n=1) |
100.0% (n=1) |
0.0% (n=1) |
100.0% (n=1) |
56.7% (n=1) |
| 男性一般保養 |
100.0% (n=1) |
0.0% (n=1) |
0.0% (n=1) |
0.0% (n=1) |
93.3% (n=1) |
| 男性備孕 |
100.0% (n=1) |
0.0% (n=1) |
0.0% (n=1) |
0.0% (n=1) |
90.0% (n=1) |
| 知識與產品查詢 |
100.0% [100.0–100.0] (n=11) |
100.0% [100.0–100.0] (n=11) |
63.6% [36.4–90.9] (n=11) |
97.0% [90.9–100.0] (n=11) |
89.4% [74.8–98.2] (n=11) |
| 詢問出貨時間 |
0.0% (n=1) |
100.0% (n=1) |
50.0% (n=1) |
66.7% (n=1) |
90.0% (n=1) |
| 詢問吃產品會胖嗎? |
100.0% (n=1) |
0.0% (n=1) |
— |
100.0% (n=1) |
93.3% (n=1) |
| 詢問退貨 |
100.0% [100.0–100.0] (n=2) |
100.0% [100.0–100.0] (n=2) |
50.0% [50.0–50.0] (n=2) |
91.7% [83.3–100.0] (n=2) |
66.7% [46.7–86.7] (n=2) |
Worst-3 cells (lowest primary score)
- 詢問吃產品會胖嗎? × Tool · 0.0% (n=1) · lowest sub_metric:
tools_recall
- 詢問出貨時間 × Scenario · 0.0% (n=1) · lowest sub_metric:
scenario_precision
- 男性備孕 × Faith · 0.0% (n=1) · lowest sub_metric:
rule_compliance