You need to agree to share your contact information to access this dataset

This repository is publicly accessible, but you have to accept the conditions to access its files and content.

Log in or Sign Up to review the conditions and access this dataset content.

tw-cyberethic

台灣繁體中文 AI 資安與倫理(cyberethic)資料集。

資料來源為本倉庫根目錄的 datasets.jsonl,共 1253 筆,每筆為一個 JSON 物件。 資料集設定(configs)已指定 train split 直接讀取 datasets.jsonl,不再使用 data/ 下的 parquet 檔。

污染修復情況(2026-08-11)

  • 修復方式:污染掃描(詞庫驅動,/data/sdg_data/sdg_pipeline/lexicon/lexicon.yaml)→ scan-triage 逐筆人工判讀「真污染 / 誤報」→ 只修真污染;簡繁/大陸用語為逐詞判讀替換,非 OpenCC 整篇轉換(避免把已正確的繁體改壞)
  • 修復範圍:僅修改 assistant 角色內容(user/tool/system 未動,訓練時本就被 mask)
  • 本次修復筆數:污染清理 75 筆
  • 完整逐筆 before/after 對照:見 cleanup_0811_before_after.jsonl;scan-triage 判讀紀錄見 cleanup_0811_triage.jsonl

補充修復(2026-08-12)

  • 修復類型:Unicode 標籤字元(U+E0000 區段)內容修正——原內容以標籤字元編碼隱藏 ASCII 指令做為資安示範教學用途,經檢視調整
  • 本次修復筆數:1 筆
  • 完整逐筆 before/after 對照:見 tagchar_fix_0812.jsonl
  • 本輪為針對性清理,非全文重新掃描,不代表已排除其他類型污染。
Downloads last month
19