兩項目評估報告
- EvalRun: #216(suite #114
auto-e2abf68d-r1-053959363)
- Target: #2 production-canary
- Cases: 28(N=1, attempts=28/28)
- Generated: 2026-05-16T13:51:09+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 / 28 (35.7%)
- 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 |
❌ 28.1% |
✅ 93.8% |
待修方向(worst-3)
- 知識與產品查詢 × retrieval_relevance = 28.1% — 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 = 93.8% — 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 |
18 |
100.0% |
| Step 1: scenario_routing |
6 |
33.3% |
| Step 2: + tool_calling |
6 |
33.3% |
| Step 3: + answer |
2 |
11.1% |
Per-scenario funnel
| Scenario (n_attempts) |
Step 1 (routing) |
Step 2 (tool) |
Step 3 (answer) |
end-to-end |
| early_return (7) |
❌ 0/7 (0.0%) |
❌ 0/7 (0.0%) |
❌ 0/7 (0.0%) |
❌ 0/7 (0.0%) |
| 產品異常 (3) |
❌ 0/3 (0.0%) |
❌ 0/3 (0.0%) |
❌ 0/3 (0.0%) |
❌ 0/3 (0.0%) |
| 知識與產品查詢 (3) |
✅ 3/3 (100.0%) |
✅ 3/3 (100.0%) |
⚪ 2/3 (66.7%) |
⚪ 2/3 (66.7%) |
| 詢問各型號重量尺寸 (5) |
⚪ 3/5 (60.0%) |
⚪ 3/5 (60.0%) |
❌ 0/5 (0.0%) |
❌ 0/5 (0.0%) |
Drop-off 最大的 5 個 scenario
- early_return drop at step1 (-100.0pp)
- 產品異常 drop at step1 (-100.0pp)
- 詢問各型號重量尺寸 drop at step3 (-60.0pp)
- 知識與產品查詢 drop at step3 (-33.3pp)
Audit
Reproduce combined report: bin/rails runner "puts Eval::EvaluationReport.call(run: EvalRun.find(216))"
Or fetch each item separately:
bin/rails runner "puts Eval::KbAccuracyReport.call(run: EvalRun.find(216))"
bin/rails runner "puts Eval::ScenarioFunnelReport.call(run: EvalRun.find(216))"
Per-Scenario × Per-Dim — Run #216
Suite: Service Client Intelligent - Rhino (bulk R1) · scenarios: 4 · dims: 5 · populated cells: 18/20
| Scenario |
Scenario |
Tool |
Retrieval |
Faith |
AnsQ |
| early_return |
0.0% [0.0–0.0] (n=7) |
0.0% [0.0–0.0] (n=2) |
83.3% [50.0–100.0] (n=6) |
100.0% [100.0–100.0] (n=5) |
75.2% [45.2–98.6] (n=7) |
| 產品異常 |
100.0% [100.0–100.0] (n=3) |
— |
0.0% [0.0–0.0] (n=3) |
100.0% [100.0–100.0] (n=2) |
94.4% [90.0–96.7] (n=3) |
| 知識與產品查詢 |
100.0% [100.0–100.0] (n=13) |
46.2% [23.1–76.9] (n=13) |
76.9% [53.8–100.0] (n=13) |
95.5% [86.4–100.0] (n=11) |
84.4% [68.7–96.7] (n=13) |
| 詢問各型號重量尺寸 |
100.0% [100.0–100.0] (n=5) |
— |
0.0% [0.0–0.0] (n=4) |
40.0% [0.0–80.0] (n=5) |
29.3% [26.7–34.7] (n=5) |
Worst-3 cells (lowest primary score)
- early_return × Scenario · 0.0% (n=7) · lowest sub_metric:
scenario_precision
- early_return × Tool · 0.0% (n=2) · lowest sub_metric:
tools_precision
- 詢問各型號重量尺寸 × Retrieval · 0.0% (n=4) · lowest sub_metric:
knowledges_source_recall