Assess the effect of 2025 legal instruments on gold-citation validity

Determine whether the 2025 legal instruments changed the rules cited by the gold answers and quantify the resulting magnitude of temporal-validity risk for system evaluation.

Background

The evaluation corpus combines the 2019 Vietnamese Labour Code, its 2026 consolidated text, and post-2024 amending and implementing instruments. Gold citations are represented only by bare article numbers, so the same citation may resolve to provisions from different legal versions.

The paper explicitly states that a provision-by-provision legal review has not been completed. Consequently, systems may be penalized for returning legally current provisions or rewarded for returning obsolete ones, and the magnitude of this evaluation bias remains unbounded. Resolving the problem requires comparing the rules cited in the gold answers against the 2025 instruments and quantifying how any changes affect retrieval and answer-scoring results.

References

We have not completed a provision-by-provision review of whether the 2025 instruments change the rules the gold answers cite (Section~\ref{sec:dataset}). Where a rule did change, the system could be penalised for returning the legally current answer, or credited for returning an obsolete one. We report the direction of this risk but cannot bound its magnitude.

— Cross-Lingual Legal QA for Vietnamese Labour Law: Retrieval, Translation, and Verifier-Guided Correction  (2609.27376 - Chi et al., 23 Sep 2026) in Limitations, paragraph “Temporal validity of gold labels” (referring to Section Dataset Construction)

Temporal drift is an unsolved problem for multi-element benchmarks.

— ListQA: A Benchmark for Evaluating List-Formatted Factual Knowledge Retrieval in Large Language Models  (2610.19165 - Kulkarni et al., 13 Oct 2026) in Section 1, Introduction, paragraph “Temporal drift is an unsolved problem for multi-element benchmarks”