From 21 to 24 September 2026, CEFC Taipei will host, in collaboration with the Huma-Num DISTAM+ consortium and with the support of the BFT, the hackathon “LLMs and the Challenge of Time”, taking place simultaneously at Academia Sinica in Taipei and Campus Condorcet in Aubervilliers.
Over three days, participants will work with a corpus specifically compiled for the occasion to assess the ability of large language models (LLMs) to extract temporal information, including chronology, dating, event duration, and temporal relationships between events.
The experiments will be conducted on texts of varying lengths in English, Standard Chinese, and Classical Chinese.
Programme
21 September — Launch of the hackathon and identification of avenues for exploration
22–23 September — Explorations and experiments conducted alternately by the Taipei and Paris teams, with joint sessions to share and discuss progress
24 September — Synthesis, report writing, documentation, and discussion of a possible publication of the results
For more information:
English version: https://distam.hypotheses.org/27287
2026 年 9 月 21 日至 24 日,法國現代中國研究中心台北辦公室(CEFC Taipei)將與 Huma-Num DISTAM+ 聯盟合作,並在 BFT 的支持下,舉辦「大型語言模型的時間挑戰」(Les LLMs au défi du temps)Hackathon 工作坊。活動將同步於台北中央研究院與法國歐貝維利耶(Aubervilliers)的 Campus Condorcet 兩地進行。
為期三天的工作坊中,參與者將以本次活動特別建置的語料庫為基礎,測試並評估大型語言模型(LLMs)擷取時間資訊的能力,包括事件的時間順序、年代判定、事件持續時間,以及不同事件之間的時間關係。
實驗將針對不同篇幅的英文、現代標準漢語及古典漢語文本進行。
活動日程
9 月 21 日 — Hackathon 啟動、共同討論並確立研究與實驗方向
9 月 22–23 日 — 台北與巴黎團隊交替進行探索與實驗,並安排共同討論與成果交流
9 月 24 日 — 工作成果統整、報告撰寫與資料整理,並討論後續成果發表的可能性

