兩項目評估報告
- EvalRun: #107(suite #53
auto-5b083fe6-r1-053958626)
- Target: #1 production-baseline
- Cases: 13(N=1, attempts=13/13)
- Generated: 2026-05-16T13:17:18+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: 11 / 13 (84.6%)
- 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 |
| early_return |
0 / 2 |
— |
— |
| 知識與產品查詢 |
11 / 11 |
❌ 0.0% |
— |
待修方向(worst-3)
- 知識與產品查詢 × retrieval_relevance = 0.0% — bot 進對 scenario 但沒呼叫 search tool — 流程 / prompt 問題(不是 KB content gap)
項目二: 情境調用與完成
整體 funnel(全 scenarios 加總)
| Stage |
Pass count |
% of total |
| Total attempts |
13 |
100.0% |
| Step 1: scenario_routing |
11 |
84.6% |
| Step 2: + tool_calling |
11 |
84.6% |
| Step 3: + answer |
3 |
23.1% |
Per-scenario funnel
| Scenario (n_attempts) |
Step 1 (routing) |
Step 2 (tool) |
Step 3 (answer) |
end-to-end |
| early_return (2) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
❌ 0/2 (0.0%) |
| 知識與產品查詢 (11) |
✅ 11/11 (100.0%) |
✅ 11/11 (100.0%) |
❌ 3/11 (27.3%) |
❌ 3/11 (27.3%) |
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 (-72.7pp)
Audit
Reproduce combined report: bin/rails runner "puts Eval::EvaluationReport.call(run: EvalRun.find(107))"
Or fetch each item separately:
bin/rails runner "puts Eval::KbAccuracyReport.call(run: EvalRun.find(107))"
bin/rails runner "puts Eval::ScenarioFunnelReport.call(run: EvalRun.find(107))"
Per-Scenario × Per-Dim — Run #107
Suite: CoolBe Agnet (bulk R1) · scenarios: 2 · dims: 5 · populated cells: 7/10
| Scenario |
Scenario |
Tool |
Retrieval |
Faith |
AnsQ |
| early_return |
0.0% [0.0–0.0] (n=2) |
0.0% [0.0–0.0] (n=2) |
— |
— |
61.7% [40.0–83.3] (n=2) |
| 知識與產品查詢 |
100.0% [100.0–100.0] (n=11) |
100.0% [100.0–100.0] (n=11) |
— |
0.0% [0.0–0.0] (n=11) |
58.5% [44.8–72.7] (n=11) |
Worst-3 cells (lowest primary score)
- early_return × Scenario · 0.0% (n=2) · lowest sub_metric:
scenario_precision
- early_return × Tool · 0.0% (n=2) · lowest sub_metric:
tools_precision
- 知識與產品查詢 × Faith · 0.0% (n=11) · lowest sub_metric:
rule_compliance