兩項目評估報告
- EvalRun: #241(suite #133
auto-16534080-r1-053959661)
- Target: #1 production-baseline
- Cases: 30(N=1, attempts=30/30)
- Generated: 2026-05-16T13:51:00+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 |
❌ 32.6% |
✅ 100.0% |
待修方向(worst-3)
- 知識與產品查詢 × retrieval_relevance = 32.6% — 撈到的 KB chunks 不夠相關 — KB content gap,跟 Neptune team 提 enrichment
- 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 |
2 |
10.0% |
| Step 2: + tool_calling |
2 |
10.0% |
| Step 3: + answer |
2 |
10.0% |
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%) |
| 修改訂單/取消訂單 (2) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
| 庫存查詢和缺貨推薦 (2) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
| 找不到退貨鈕/退貨按鍵/退貨申請 (2) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
| 沒用到優惠/折扣想補用 (2) |
⚪ 1/2 (50.0%) |
⚪ 1/2 (50.0%) |
⚪ 1/2 (50.0%) |
⚪ 1/2 (50.0%) |
| 詢問出貨/到貨時間 (1) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
❌ 0/1 (0.0%) |
| 退貨/換貨 (2) |
⚪ 1/2 (50.0%) |
⚪ 1/2 (50.0%) |
⚪ 1/2 (50.0%) |
⚪ 1/2 (50.0%) |
Drop-off 最大的 5 個 scenario
- early_return drop at step1 (-100.0pp)
- 修改訂單/取消訂單 drop at step1 (-100.0pp)
- 庫存查詢和缺貨推薦 drop at step1 (-100.0pp)
- 找不到退貨鈕/退貨按鍵/退貨申請 drop at step1 (-100.0pp)
- 詢問出貨/到貨時間 drop at step1 (-100.0pp)
Audit
Reproduce combined report: bin/rails runner "puts Eval::EvaluationReport.call(run: EvalRun.find(241))"
Or fetch each item separately:
bin/rails runner "puts Eval::KbAccuracyReport.call(run: EvalRun.find(241))"
bin/rails runner "puts Eval::ScenarioFunnelReport.call(run: EvalRun.find(241))"
Per-Scenario × Per-Dim — Run #241
Suite: VERVE 小幫手 (bulk R1) · scenarios: 8 · dims: 5 · populated cells: 35/40
| 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=8) |
89.3% [76.7–99.3] (n=9) |
| 修改訂單/取消訂單 |
100.0% [100.0–100.0] (n=2) |
— |
0.0% [0.0–0.0] (n=2) |
100.0% [100.0–100.0] (n=2) |
100.0% [100.0–100.0] (n=2) |
| 庫存查詢和缺貨推薦 |
100.0% [100.0–100.0] (n=2) |
100.0% [100.0–100.0] (n=2) |
— |
0.0% [0.0–0.0] (n=2) |
36.7% [26.7–46.7] (n=2) |
| 找不到退貨鈕/退貨按鍵/退貨申請 |
0.0% [0.0–0.0] (n=2) |
— |
100.0% (n=1) |
100.0% [100.0–100.0] (n=2) |
90.0% [90.0–90.0] (n=2) |
| 沒用到優惠/折扣想補用 |
100.0% [100.0–100.0] (n=2) |
— |
0.0% [0.0–0.0] (n=2) |
0.0% [0.0–0.0] (n=2) |
96.7% [96.7–96.7] (n=2) |
| 知識與產品查詢 |
100.0% [100.0–100.0] (n=10) |
0.0% [0.0–0.0] (n=10) |
90.0% [70.0–100.0] (n=10) |
100.0% [100.0–100.0] (n=9) |
97.3% [94.3–100.0] (n=10) |
| 詢問出貨/到貨時間 |
100.0% (n=1) |
— |
0.0% (n=1) |
100.0% (n=1) |
100.0% (n=1) |
| 退貨/換貨 |
50.0% [0.0–100.0] (n=2) |
50.0% [0.0–100.0] (n=2) |
75.0% [50.0–100.0] (n=2) |
100.0% [100.0–100.0] (n=2) |
58.3% [26.7–90.0] (n=2) |
Worst-3 cells (lowest primary score)
- early_return × Scenario · 0.0% (n=9) · lowest sub_metric:
scenario_precision
- early_return × Tool · 0.0% (n=1) · lowest sub_metric:
tools_precision
- 詢問出貨/到貨時間 × Retrieval · 0.0% (n=1) · lowest sub_metric:
knowledges_source_recall