Beyond Known Facts: Generating Unseen Temporal Knowledge to Address Data Contamination in LLM Evaluation CCS: Computing methodologies Information extraction CCS: Information systems Web mining CCS: Computing methodologies Language resources

Arthur Amalvy, Hen-Hsen Huang · 2026

Read Beyond Known Facts: Generating Unseen Temporal Knowledge to Address Data Contamination in LLM Evaluation CCS: Computing methodologies Information extraction CCS: Information systems Web mining CCS: Computing methodologies Language resources by Arthur Amalvy, Hen-Hsen Huang (2026) 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-2601-13658.txt.

Description

Arthur Amalvy and Hen-Hsen Huang work on pulling time-stamped facts from text for the web in 2026. Big knowledge bases like Wikidata need facts that say when things held true. Temporal knowledge graph extraction pulls such facts as quadruples: subject, relation, object, and time. The hard part is facts not seen in training data.

Their work, framed by its title as going beyond known facts, makes new unseen temporal knowledge. Fresh facts help test large language models without data contamination. Contamination occurs when test items leaked into training data. Clean tests give a truer read on what models know.

Catalog record

CREATORArthur Amalvy, Hen-Hsen Huang
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE51 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2601-13658

Browse the full library · About the Tower of Babel