lianghsun/tw-legal-benchmark-v2

Taiwan Legal Benchmark v2

A multiple-choice benchmark for evaluating LLMs on Taiwan law in Traditional Chinese, built from 15 years (2012–2026) of national examinations published by the Ministry of Examination (考選部).

Supersedes tw-legal-benchmark-v1 (209 questions) with 17,002 deduplicated questions across 15 legal domains.

Overview

Property	Value
Questions	17,002 (deduplicated)
Years	2012–2026
Source papers	1,040 official exam papers
Format	4-choice single-answer (A/B/C/D)
Language	Traditional Chinese (zh-TW)
Framework	Twinkle Eval / inspect-ai compatible

Two tiers

The benchmark is explicitly stratified by how law-dense the source paper is, because Taiwan's national examinations differ enormously in legal depth.

Tier	Questions	Source
1 — core legal	4,864	綜合法學 / 基礎法學 — the Bar and Judicial Examination first stage, the most law-dense multiple-choice examination in Taiwan
2 — domain regulation	12,139	Subject-specific statutory papers: administrative law, immigration, pharmaceutical, coastguard, construction, tax, corrections, postal, indigenous affairs

Filter with tier == 1 for the hardest, most purely legal subset.

Tier 1 is near-complete coverage of its population. Across 2012–2026 the Ministry of Examination published only 177 core-legal multiple-choice papers in total; this dataset draws on 172 of them (97%). Taiwan's senior legal subjects are overwhelmingly examined by essay, not multiple choice, so the pool of serious legal MCQs is small by construction — not by sampling.

Subject Distribution

Subject	Questions	Coverage
admin	4,270	行政法
criminal	1,847	刑法/刑事訴訟法/法律倫理
immigration	1,682	入出國及移民法規
civil	1,280	民法/民事訴訟法
public	1,205	憲法/行政法/國際公法/國際私法
pharma	1,172	藥事行政與法規
coastguard	1,143	海巡法規
construction	1,090	營建法規
commercial	930	公司法/保險法/票據法/證券交易法等
tax	596	稅務相關法規
accounting	594	記帳相關法規
corrections	397	監獄行刑法
indigenous	299	原住民族行政及法規
foundational	250	民法/刑法/行政法(基礎法學)
postal	248	郵政法規

Examination Levels

Level	Questions
四等考試	6,554
三等考試	4,403
專技高考	2,823
專技普考	1,190
普通考試	1,136
佐級晉員級	548
員級考試	249
初等考試	100

Fields

Field	Description
question	Question stem
A / B / C / D	Option texts
answer	Gold option letter
choices	JSON array of the four options (inspect-ai compatible)
year / year_roc	Exam year (CE / ROC calendar)
subject / subject_zh	Subject code and Chinese label
tier	1 = core legal, 2 = domain regulation
exam_level / exam_name	Examination grade and official title
q_no	Question number on the original paper
source_papers / n_source_papers	Originating paper id(s) and multiplicity

Usage

Twinkle Eval

uv run scripts/make_eval_config.py \
    --model my-model --base-url http://localhost:8000/v1 \
    --bench lianghsun/tw-legal-benchmark-v2 --out eval/
twinkle-eval --config eval/config.yaml --export json csv

If a model scores near zero, check the 無法解析 rate before believing the number. A stock instruct model answering 最終答案：C scores ~5% under the box method and ~30% under pattern — same model, same questions. Match the extraction method to the model.

datasets

from datasets import load_dataset
ds = load_dataset("lianghsun/tw-legal-benchmark-v2", split="train")
core = ds.filter(lambda r: r["tier"] == 1)      # hardest subset

Construction

1. Index — the official MOEX open-data index (wHandExamQandA_CSV.ashx, 66,582 rows), filtered to statute-grounded multiple-choice papers: 1,040 papers.
2. Retrieval — 2,080 question and answer-key PDFs fetched (rate-limited, checksummed, 0 failures).
3. Parsing — 986/1,040 papers parsed completely; 56,479 raw questions.
4. Filtering and deduplication — see below.

Parsing hazards handled

These silently corrupt naive extraction, so they are documented here:

• Option markers live in the EUDC Private Use Area (U+E18C–U+E190 = A–E) rather than as literal letters — and one legacy layout uses plain 1. / A. instead.
• Text mixes CJK Compatibility Ideographs (e.g. 不 U+F967). Normalised with NFC — not NFKC, which rewrites 「，」to an ASCII comma and breaks matching against statute text.
• Papers mix single-select and multi-select sections; multi-select answers are multi-letter tokens (ABDE) that a per-character regex shreds, misaligning every later row.
• Answer sheets appear in several layouts: 01 vs 第1題 numbering, 題號 vs 題序 labels, half-width vs full-width answer letters (Ａ, ＃).
• Variable spacing after question numbers (one paper year uses a single space).
• Nine 2014 answer sheets carry an internally inconsistent official range (「單選題數：70題(第169題)」 overlapping 複選 6170).

Deduplication

The same paper is republished under several examination categories (司法官 / 律師 / 司法官及律師, and equivalents at other grades). Before deduplication there are 56,095 single-answer items; after collapsing identical items 17,002 remain. All duplicate groups were verified to carry identical answer keys (0 conflicts). n_source_papers records the multiplicity.

A benchmark built from this source without deduplication is inflated roughly 2–3×, and accuracy is skewed toward the republished subset.

Excluded

Excluded	Reason
法學知識與英文 / 法學大意 papers	Shallow general-knowledge items, partly English; not statute-grounded
法學英文 sections (645 items)	Inside 綜合法學(二) papers, these test English comprehension, not Taiwanese law. Removed by script-ratio detection, not by question number
Multi-select items	Different task format
Voided questions (一律給分)	No valid answer exists
Multiple-accepted-answer items	Answer key not single-valued
52 papers with no extractable option markers	PDF text layer lacks the glyphs; excluded rather than guessed
2 papers that parsed incompletely	Risk of silently merged stems

Known Properties

• Answer-position bias in the source: A=3,521, B=4,334, C=4,639, D=4,509. Use shuffle_options: true (Twinkle Eval default).
• Contamination risk: these are public past papers, widely discussed online and very likely present in pretraining corpora. Absolute scores should be treated with caution; the benchmark is most useful for relative comparison between models.
• Whitespace normalisation removes spaces inside stems, which is correct for Chinese and would be wrong for Latin text — the English sections are excluded for this reason too.
• Item difficulty is set by the examination committee, not by us.

Licensing

Questions and answer keys are official publications of the Ministry of Examination, Republic of China (Taiwan), released for public access. The packaging, parsing, deduplication, and metadata in this repository are provided under Apache-2.0, following the convention of v1. Users intending redistribution should confirm the current terms published by 考選部.

Citation

@misc{twlegalbenchv2,
  title  = {Taiwan Legal Benchmark v2},
  author = {Huang, Liang-Hsun},
  year   = {2026},
  url    = {https://huggingface.co/datasets/lianghsun/tw-legal-benchmark-v2}
}