The Flow has not summarised this story yet — read it at arXiv Computation and Language.
The paper introduces SimpCue, a cue-based prompting approach for multilingual text simplification in Catalan, Spanish, and Italian. It compares three prompting strategies—baseline, gold-cue (with gold linguistic cues), and predicted-cue (with automatically predicted cues)—using the Qwen3-8B model. Results show that predicted-cue prompting yields the best overall scores across SARI, BLEU, chrF, and BERTScore, though gains over the baseline are modest and vary by language.
The paper introduces a method for text simplification that uses a Large Language Model-as-a-Judge (LLM‑Judge) to automatically generate policy‑aligned training data, eliminating the need for costly human annotation or parallel corpora. This approach allows simplification models to adapt to various policies—such as lexical‑focused simplification or overall rewriting that trades detail for simplicity—at both sentence and document levels. Experiments show that small‑scale open‑source LLMs can match or surpass GPT‑5, with consistent improvements across model families and sizes.
arXiv:2603.02041v2 Announce Type: replace-cross Abstract: Large language models (LLMs) are predominantly trained on English-centric data, resulting in uneven performance for smaller languages. We stu...
The paper presents a pipeline that leverages large language models to extract grammatical rules, example sentences, and lexicons from descriptive grammar books, producing synthetic parallel corpora for fine‑tuning machine translation models. Evaluated on three low‑resource languages—Kalamang, Tuatschin, and Mandan—the synthetic data improves translation quality over seed‑data baselines in 75% of configurations for Kalamang and 59% for Tuatschin, achieving up to +8.8 ChrF++ gains. A factorial study across 96 configurations identifies which combinations of target part‑of‑speech, retrieval granularity, and sample volume drive performance gains and where they fail, demonstrating that static linguistic documentation can be repurposed for practical translation tools for severely under‑resourced languages.
Agentic benchmarks aim to measure how well AI agents plan, search, execute, and recover within realistic multi-tool environments, but they are almost exclusively in English. As AI agents are globally deployed to a linguistically diverse user base, whether agentic competence measured in English transfers to other languages remains an open question.
arXiv:2609.18284v1 Announce Type: new Abstract: In recent years, three initiatives have emerged to develop generative language models in Hungary. The motivation behind them is the same. For Hungarian...