arXiv cs.CL· Rik van Noord, Maurice Voors·· 23 小时前AI 评分29
The Shared Word(TSW-NL):评估语言模型在荷兰语字谜上的表现
The Shared Word: Evaluating Language Models on Dutch Word Puzzles
AI 导读
研究者构建了荷兰语基准 TSW-NL,用三条指向同一荷兰语词不同义项的歧义线索考察模型跨线索整合能力,共 222 道题、每题含三名独立人类作答。人类得分区间为 49.5%–61.0%,测试中最高分开源权重模型仅 39.3%,多数模型低于 20%。闭源模型表现突出:GPT-4.1(2025 年 4 月)得 59.2%,GPT-5.6 Sol 开启最大推理后达 88.8%。
正文 · 原文
Abstract:Many LLM benchmarks for languages other than English are translated, multiple-choice, or difficult to score automatically. We introduce The Shared Word (TSW-NL), a Dutch benchmark built around four core principles: it is language-specific, open-vocabulary, human-curated, and deterministically scored. Each puzzle gives three clues that describe different senses or uses of the same Dutch word. The clues are individually ambiguous, requiring the solver to combine evidence across multiple clues. We construct the benchmark from Dutch Wikipedia disambiguation pages, followed by automatic filtering and extensive manual curation. The final dataset contains 222 instances, with three independent human answers per test item. The task is clearly challenging: scores in our human study range from 49.5% to 61.0% and the highest-scoring open-weight model we tested reaches only 39.3%. Many others even score below 20%. At the same time, closed-source models show great progress: GPT-4.1 (April 2025) scores 59.2%, whereas GPT-5.6 Sol with max reasoning scores 88.8%. We believe this task provides a useful addition to Dutch model evaluation, plus a clear recipe that can be applied to construct a similar benchmark for other languages.
| Comments: | EACL 2027 (Findings) |
| Subjects: | Computation and Language (cs.CL) |
| Cite as: | arXiv:2611.22935 [cs.CL] |
| (or arXiv:2611.22935v1 [cs.CL] for this version) | |
| https://doi.org/10.48550/arXiv.2611.22935 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Rik Van Noord [view email]
[v1]
Tue, 17 Nov 2026 15:30:26 UTC (1,489 KB)
来源:arXiv cs.CL · arxiv.org