兩項目評估報告
- EvalRun: #189(suite #102
auto-99569222-r1-053959111)
- Target: #1 production-baseline
- Cases: 30(N=1, attempts=30/30)
- Generated: 2026-05-16T13:52:22+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 |
❌ 44.8% |
✅ 100.0% |
待修方向(worst-3)
- 知識與產品查詢 × retrieval_relevance = 44.8% — 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 |
3 |
15.0% |
| Step 2: + tool_calling |
3 |
15.0% |
| 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%) |
| 建議轉接 (3) |
❌ 1/3 (33.3%) |
❌ 1/3 (33.3%) |
❌ 1/3 (33.3%) |
❌ 1/3 (33.3%) |
| 消費者補點 (4) |
❌ 0/4 (0.0%) |
❌ 0/4 (0.0%) |
❌ 0/4 (0.0%) |
❌ 0/4 (0.0%) |
| 發票照片辨識 (2) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
| 知識與產品查詢 (2) |
✅ 2/2 (100.0%) |
✅ 2/2 (100.0%) |
⚪ 1/2 (50.0%) |
⚪ 1/2 (50.0%) |
Drop-off 最大的 5 個 scenario
- early_return drop at step1 (-100.0pp)
- 消費者補點 drop at step1 (-100.0pp)
- 發票照片辨識 drop at step1 (-100.0pp)
- 建議轉接 drop at step1 (-66.7pp)
- 知識與產品查詢 drop at step3 (-50.0pp)
Audit
Reproduce combined report: bin/rails runner "puts Eval::EvaluationReport.call(run: EvalRun.find(189))"
Or fetch each item separately:
bin/rails runner "puts Eval::KbAccuracyReport.call(run: EvalRun.find(189))"
bin/rails runner "puts Eval::ScenarioFunnelReport.call(run: EvalRun.find(189))"
Per-Scenario × Per-Dim — Run #189
Suite: Ocard (bulk R1) · scenarios: 5 · dims: 5 · populated cells: 22/25
| Scenario |
Scenario |
Tool |
Retrieval |
Faith |
AnsQ |
| early_return |
0.0% [0.0–0.0] (n=9) |
— |
100.0% [100.0–100.0] (n=9) |
100.0% [100.0–100.0] (n=9) |
92.2% [78.9–99.6] (n=9) |
| 建議轉接 |
100.0% [100.0–100.0] (n=3) |
— |
— |
33.3% [0.0–100.0] (n=3) |
72.2% [40.0–90.0] (n=3) |
| 消費者補點 |
100.0% [100.0–100.0] (n=4) |
25.0% [0.0–75.0] (n=4) |
0.0% [0.0–0.0] (n=4) |
100.0% [100.0–100.0] (n=4) |
68.3% [53.3–88.3] (n=4) |
| 發票照片辨識 |
0.0% [0.0–0.0] (n=2) |
0.0% [0.0–0.0] (n=2) |
0.0% [0.0–0.0] (n=2) |
0.0% [0.0–0.0] (n=2) |
26.7% [26.7–26.7] (n=2) |
| 知識與產品查詢 |
100.0% [100.0–100.0] (n=12) |
100.0% [100.0–100.0] (n=12) |
90.9% [72.7–100.0] (n=11) |
94.4% [83.3–100.0] (n=12) |
93.3% [86.1–98.9] (n=12) |
Worst-3 cells (lowest primary score)
- early_return × Scenario · 0.0% (n=9) · lowest sub_metric:
scenario_precision
- 發票照片辨識 × Faith · 0.0% (n=2) · lowest sub_metric:
rule_compliance
- 發票照片辨識 × Retrieval · 0.0% (n=2) · lowest sub_metric:
knowledges_source_precision