🤖 AI Summary
This study addresses the “temporal misalignment” problem in legal Retrieval-Augmented Generation (RAG), which arises when statutory provisions are treated without regard to their temporal validity. The work provides the first systematic definition and quantification of this issue, introduces a multi-version corpus spanning 93 years of French tax law, and presents FiscalQA Pro—the first version-aware legal question-answering benchmark. The proposed approach formulates legal QA as a time-indexed retrieval task, leveraging multi-version indexing, an atomic scoring mechanism based on regular expressions and numerical tolerance, and an end-to-end learnable retriever. Crucially, it avoids reliance on large language models for judgment to prevent temporal bias. Experiments show that static RAG achieves 0% recall for the correct statutory version, whereas the proposed method attains a strict accuracy of 98.3% without oracle information, approaching the oracle upper bound of 99.1%.
📝 Abstract
We identify and quantify temporal misgrounding: the systematic retrieval and citation of the currently in-force version of a legal article when the applicable version is an earlier or future one. Standard legal RAG treats the corpus as static; we argue legal question answering is a temporally-indexed retrieval problem. We introduce FiscalQA Pro, pairing a versioned corpus of 32,436 article-versions of the French tax code (93 years, 1938-2031) with an all-model-hard temporal-reasoning track: 209 scored, expert-reviewed questions across 33 CGI articles (221 released; twelve flagged out of the answerable scope). At selection time, no evaluated model recovered its date-applicable answer closed-book in any of four sampling draws, and the currently in-force text lacks the gold value for all but one of the scored questions. Answers are scored deterministically via atomic ground-truth "nuggets" (regex and numeric-with-tolerance), never LLM-as-judge: an LLM judge would inherit the temporal bias it is meant to score. Across eleven models (five frontier closed-API systems plus Gemini 2.5 Pro as a substitute entry, and five open-weight), parametric knowledge yields 3.0% mean strict accuracy and RAG over a static current-version corpus 2.7%. Static RAG retrieves the date-applicable version 0% of the time, confidently citing a real but inapplicable version. Our end-to-end retriever over a multi-version index, with no oracle, reaches 98.3% mean strict; an oracle-article ablation reaches 99.1%, locating the residual gap in first-stage recall, not version selection. We additionally release a version-aware jurisprudence dataset of 69,208 citation links, together with the corpus, benchmark, model responses, and pipeline code.