第 21 名
Loverse
60.6C/O 分數
示意預覽：在此測試環境資料集中，loverse 於「敘事主動性」表現最佳，於「延遲回想」表現最弱（方法論 v0.1）。
兩項重大弱點
- 示意紀錄：「延遲回想」在測試環境情境集中得分 47／100。
- 示意紀錄：「角色一致性」在測試環境情境集中得分 49／100。
適合哪些讀者
- 適合的情況
- 適合把「敘事主動性」看得比其他評分項目更重要的讀者。
- 不適合的情況
- 不適合需要正式驗證證據、而非測試環境預覽的讀者。
測試項目
| 項目 | 結果 |
|---|---|
| Conversation quality | 61 |
| Local language and culture | 68 |
| Continuity and memory | 57.4 |
| Price and value | 61.5 |
| Product usability and reliability | 63 |
| Character and roleplay | 63.8 |
分類分數
| 分類 | C/O 分數 |
|---|---|
| 總體 | 60.6 |
| 關係 | 59 |
| 角色扮演 | 61.5 |
| 記憶 | 56.5 |
| 性價比 | 61.3 |
記憶時間軸
- continuity 57/1002026-08-15
- contradiction-resistance 57/1002026-08-15
- delayed-recall 47/1002026-08-15
- memory-controls 54/1002026-08-15
- relationship-continuity 72/1002026-08-15
當地語言觀察
- local-language-performance: 68/100 (Illustrative staging scenario result recorded on the published 0-100 scale.)
價格與付費牆觀察
| 測試的方案 | paid-standard |
|---|---|
| 觀測價格 | NT$490.00 |
| 免費額度 | 26 messages per day |
| 觀測日期 | 2026-08-12 |
三個替代方案
- Nomi 78.9
- Character.AI 78.3
- Kindroid 77.4
方法與日期
依方法 v0.1 於 2026-08-24 計算。
更正
以下為紀錄中保存的英文原文，未經翻譯。
- 2026-08-24 app: apps — A destination may only be published once it has been resolved, not assumed. (Edition-specific official destinations assumed → Each edition points at the same verified official entry point until Task 7 verifies edition-specific destinations; 分數沒有變動)
- 2026-08-23 price: prices/kr — Price facts must carry the date they were observed so staleness is visible. (Korean tier prices recorded without a stated observation date → Every Korean price observation carries an explicit 2026-08-12 observation date; 分數沒有變動)
- 2026-08-22 methodology: 0.1 — The published weight table needed an unambiguous definition before any score was calculated. (Overall product/local-language component described as a single dimension → Overall product/local-language component is the mean of product reliability and local-language performance; 分數有變動)