兩項目評估報告
- EvalRun: #80（suite #44
auto-97fef389-r1-053958491）
- Target: #2 production-canary
- Cases: 30（N=1, attempts=30/30）
- Generated: 2026-05-16T12:25:25+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 |
⚪ 57.9% |
✅ 100.0% |
待修方向（worst-3）
- 知識與產品查詢 × retrieval_relevance = 57.9% — 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 |
8 |
40.0% |
| Step 2: + tool_calling |
8 |
40.0% |
| Step 3: + answer |
5 |
25.0% |
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%) |
| 優惠活動-登入發票送全家點數/聖誕交易挑戰 (1) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
| 疑似詐騙 (1) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
✅ 1/1 (100.0%) |
| 知識與產品查詢 (4) |
✅ 4/4 (100.0%) |
✅ 4/4 (100.0%) |
❌ 1/4 (25.0%) |
❌ 1/4 (25.0%) |
| 訂單查詢 (3) |
❌ 0/3 (0.0%) |
❌ 0/3 (0.0%) |
❌ 0/3 (0.0%) |
❌ 0/3 (0.0%) |
| 驗證 (2) |
✅ 2/2 (100.0%) |
✅ 2/2 (100.0%) |
✅ 2/2 (100.0%) |
✅ 2/2 (100.0%) |
Drop-off 最大的 5 個 scenario
- early_return drop at step1 (-100.0pp)
- 訂單查詢 drop at step1 (-100.0pp)
- 知識與產品查詢 drop at step3 (-75.0pp)
- 優惠活動-登入發票送全家點數/聖誕交易挑戰 drop at step1 (-0.0pp)
- 疑似詐騙 drop at step1 (-0.0pp)
Audit
Reproduce combined report: bin/rails runner "puts Eval::EvaluationReport.call(run: EvalRun.find(80))"
Or fetch each item separately:
bin/rails runner "puts Eval::KbAccuracyReport.call(run: EvalRun.find(80))"bin/rails runner "puts Eval::ScenarioFunnelReport.call(run: EvalRun.find(80))"
Per-Scenario × Per-Dim — Run #80
Suite: BitoPro (bulk R1) · scenarios: 6 · dims: 5 · populated cells: 27/30
| Scenario |
Scenario |
Tool |
Retrieval |
Faith |
AnsQ |
| early_return |
0.0% [0.0–0.0] (n=9) |
100.0% (n=1) |
75.0% [37.5–100.0] (n=8) |
100.0% [100.0–100.0] (n=9) |
85.2% [73.3–94.1] (n=9) |
| 優惠活動-登入發票送全家點數/聖誕交易挑戰 |
100.0% (n=1) |
0.0% (n=1) |
0.0% (n=1) |
— |
100.0% (n=1) |
| 疑似詐騙 |
100.0% (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=14) |
100.0% [100.0–100.0] (n=14) |
90.0% [70.0–100.0] (n=10) |
95.2% [91.7–98.8] (n=14) |
81.9% [66.2–93.6] (n=14) |
| 訂單查詢 |
100.0% [100.0–100.0] (n=3) |
100.0% [100.0–100.0] (n=3) |
— |
100.0% [100.0–100.0] (n=3) |
71.1% [50.0–90.0] (n=3) |
| 驗證 |
100.0% [100.0–100.0] (n=2) |
100.0% [100.0–100.0] (n=2) |
50.0% [50.0–50.0] (n=2) |
66.7% [66.7–66.7] (n=2) |
91.7% [90.0–93.3] (n=2) |
Worst-3 cells (lowest primary score)
- early_return × Scenario · 0.0% (n=9) · lowest sub_metric:
scenario_precision
- 優惠活動-登入發票送全家點數/聖誕交易挑戰 × Tool · 0.0% (n=1) · lowest sub_metric:
tools_recall
- 優惠活動-登入發票送全家點數/聖誕交易挑戰 × Retrieval · 0.0% (n=1) · lowest sub_metric:
knowledges_source_recall