# PolERo: Studying Political Evasion in Romanian

arXiv: https://arxiv.org/abs/2609.02391 · HTML: https://mlsanity.com/paper/2609.02391/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 21 of its 47 references that are in the map.
- First posted on arXiv 2 Sep 2026.

## Builds on (most cited first, 21 of 21)

- [BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding](https://mlsanity.com/paper/1810.04805/) (2018)
- [RoBERTa: A Robustly Optimized BERT Pretraining Approach](https://mlsanity.com/paper/1907.11692/) (2019)
- [Unsupervised Cross-lingual Representation Learning at Scale](https://mlsanity.com/paper/1911.02116/) (2019)
- [Know What You Don't Know: Unanswerable Questions for SQuAD](https://mlsanity.com/paper/1806.03822/) (2018)
- [Smarter, Better, Faster, Longer: A Modern Bidirectional Encoder for Fast, Memory Efficient, and Long Context Finetuning and Inference](https://mlsanity.com/paper/2412.13663/) (2024)
- [TyDi QA: A Benchmark for Information-Seeking Question Answering in Typologically Diverse Languages](https://mlsanity.com/paper/2003.05002/) (2020)
- [AmbigQA: Answering Ambiguous Open-domain Questions](https://mlsanity.com/paper/2004.10645/) (2020)
- [xCOMET: Transparent Machine Translation Evaluation through Fine-grained Error Detection](https://mlsanity.com/paper/2310.10482/) (2023)
- [The birth of Romanian BERT](https://mlsanity.com/paper/2009.08712/) (2020)
- [Did they answer? Subjective acts and intents in conversational discourse](https://mlsanity.com/paper/2104.04470/) (2021)
- [Representativeness as a Forgotten Lesson for Multilingual and Code-switched Data Collection and Preparation](https://mlsanity.com/paper/2310.20470/) (2023)
- [Introducing RONEC -- the Romanian Named Entity Corpus](https://mlsanity.com/paper/1909.01247/) (2019)
- ["I Never Said That": A dataset, taxonomy and baselines on response clarity classification](https://mlsanity.com/paper/2409.13879/) (2024)
- [Detecting Data Contamination in LLMs via In-Context Learning](https://mlsanity.com/paper/2510.27055/) (2025)
- [OpenLLM-Ro -- Technical Report on Open-source Romanian LLMs](https://mlsanity.com/paper/2405.07703/) (2024)
- [RoQLlama: A Lightweight Romanian Adapted Language Model](https://mlsanity.com/paper/2410.04269/) (2024)
- [SemEval-2026 Task 6: CLARITY -- Unmasking Political Question Evasions](https://mlsanity.com/paper/2603.14027/) (2026)
- [Target Based Speech Act Classification in Political Campaign Text](https://mlsanity.com/paper/1905.07856/) (2019)
- [A Large-Scale Benchmark for Evaluating Large Language Models on Medical Question Answering in Romanian](https://mlsanity.com/paper/2508.16390/) (2025)
- [Knowing What's Missing: Assessing Information Sufficiency in Question Answering](https://mlsanity.com/paper/2512.06476/) (2025)
- [SG-UniBuc-NLP at SemEval-2026 Task 6: Multi-Head RoBERTa with Chunking for Long-Context Evasion Detection](https://mlsanity.com/paper/2604.26375/) (2026)

## Abstract (arXiv)

Political evasion refers to responses that engage with a question while withholding the requested information. Recent NLP work frames political evasion as a classification task using a two-level taxonomy of response clarity and fine-grained evasion strategies. Existing work on response clarity and evasion classification is limited to English, leaving open whether the taxonomy and model behavior transfer across languages and political contexts. We introduce PolERo, a dataset of 3,574 human-annotated question-answer pairs extracted from official transcripts of five Romanian presidents. We evaluate multiple classification approaches on both datasets under matched conditions, including TF-IDF baselines, fine-tuned encoder models, a proposed sliding-window encoder, and zero/few-shot LLM prompting. We study cross-lingual transfer through joint bilingual training and machine-translation-based data augmentation. Our results indicate that fine-tuned encoders are competitive, cross-lingual transfer is asymmetric, and ambivalent evasion categories involving pragmatic cues remain the main challenge across all model families.