兩項目評估報告
- EvalRun: #176（suite #95
auto-2ebddf81-r1-053959053）
- Target: #2 production-canary
- Cases: 20（N=1, attempts=20/20）
- Generated: 2026-05-16T13:29:11+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: 5 / 20 (25.0%)
- 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 |
5 / 5 |
❌ 30.7% |
✅ 100.0% |
待修方向（worst-3）
- 知識與產品查詢 × retrieval_relevance = 30.7% — 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 |
15 |
100.0% |
| Step 1: scenario_routing |
7 |
46.7% |
| Step 2: + tool_calling |
7 |
46.7% |
| Step 3: + answer |
1 |
6.7% |
Per-scenario funnel
| Scenario (n_attempts) |
Step 1 (routing) |
Step 2 (tool) |
Step 3 (answer) |
end-to-end |
| early_return (4) |
❌ 0/4 (0.0%) |
❌ 0/4 (0.0%) |
❌ 0/4 (0.0%) |
❌ 0/4 (0.0%) |
| 知識與產品查詢 (5) |
✅ 5/5 (100.0%) |
✅ 5/5 (100.0%) |
❌ 0/5 (0.0%) |
❌ 0/5 (0.0%) |
| 訂單查詢 (6) |
❌ 2/6 (33.3%) |
❌ 2/6 (33.3%) |
❌ 1/6 (16.7%) |
❌ 1/6 (16.7%) |
Drop-off 最大的 5 個 scenario
- early_return drop at step1 (-100.0pp)
- 知識與產品查詢 drop at step3 (-100.0pp)
- 訂單查詢 drop at step1 (-66.7pp)
Audit
Reproduce combined report: bin/rails runner "puts Eval::EvaluationReport.call(run: EvalRun.find(176))"
Or fetch each item separately:
bin/rails runner "puts Eval::KbAccuracyReport.call(run: EvalRun.find(176))"bin/rails runner "puts Eval::ScenarioFunnelReport.call(run: EvalRun.find(176))"
Per-Scenario × Per-Dim — Run #176
Suite: MatrixCloud AI Agent (bulk R1) · scenarios: 3 · dims: 5 · populated cells: 14/15
| Scenario |
Scenario |
Tool |
Retrieval |
Faith |
AnsQ |
| early_return |
0.0% [0.0–0.0] (n=4) |
100.0% (n=1) |
100.0% [100.0–100.0] (n=3) |
100.0% [100.0–100.0] (n=4) |
71.7% [41.7–90.0] (n=4) |
| 知識與產品查詢 |
100.0% [100.0–100.0] (n=10) |
100.0% [100.0–100.0] (n=10) |
100.0% [100.0–100.0] (n=5) |
53.3% [23.3–83.3] (n=10) |
59.3% [40.0–79.3] (n=10) |
| 訂單查詢 |
66.7% [33.3–100.0] (n=6) |
50.0% [16.7–83.3] (n=6) |
— |
100.0% [100.0–100.0] (n=6) |
67.2% [48.9–84.4] (n=6) |
Worst-3 cells (lowest primary score)
- early_return × Scenario · 0.0% (n=4) · lowest sub_metric:
scenario_precision
- 訂單查詢 × Tool · 50.0% (n=6) · lowest sub_metric:
tools_precision
- 知識與產品查詢 × Faith · 53.3% (n=10) · lowest sub_metric:
rule_compliance