Temporal Information Retrieval via Time-Specifier Model Merging
SeungYoon Han, Taeho Hwang, Sukmin Cho, Soyeong Jeong, Hoyun Song, Huije Lee, Jong C. Park · 2025
Read Temporal Information Retrieval via Time-Specifier Model Merging by SeungYoon Han, Taeho Hwang, Sukmin Cho, Soyeong Jeong, Hoyun Song, Huije Lee, Jong C. Park (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2507-06782.txt.
Description
Ask a search engine for documents 'in 2015' and it often ignores the date. SeungYoon Han and co-authors fix that failure without dulling ordinary search. Their method, time-specifier model merging, trains separate retrievers for separate time expressions. Each retriever masters one kind of time phrase. The method then merges them into a single model.
The merging trick answers a known trap. Teaching a retriever temporal skill usually erodes its general skill. Across temporal and non-temporal test sets, the merged model sharpened results on date-bound queries. It also held its strength on plain ones, beating the baselines on both fronts. The released code makes the recipe easy to reuse. Archives, news search, and any collection where when matters as much as what can all use it.
Catalog record
| CREATOR | SeungYoon Han, Taeho Hwang, Sukmin Cho, Soyeong Jeong, Hoyun Song, Huije Lee, Jong C. Park |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 44 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2507-06782