Papers by Jindřich Libovický
Multilingual Text-to-Image Generation Magnifies Gender Stereotypes (2025.acl-long)
Copied to clipboard
Felix Friedrich, Katharina Hämmerl, Patrick Schramowski, Manuel Brack, Jindřich Libovický, Alexander Fraser, Kristian Kersting
| Challenge: | Text-to-image (T2I) generation models have great results in image quality, flexibility, and text alignment, but they suffer from substantial gender bias. |
| Approach: | They propose a benchmark to study gender bias in multilingual T2I models . they use multilingual prompts to account for grammatical differences influencing gender . |
| Outcome: | The proposed benchmark shows strong gender biases and language-specific differences across models. |
Beyond Literal Token Overlap: Token Alignability for Multilinguality (2025.naacl-short)
Copied to clipboard
| Challenge: | Existing studies have shown that token overlap is a strong predictor of multilinguality and cross-lingual knowledge transfer between languages with different scripts. |
| Approach: | They propose a subword token alignability metric to understand the impact and quality of multilingual tokenisation. |
| Outcome: | The proposed metric predicts multilinguality much better when scripts are disparate and the overlap of literal tokens is low. |
Lexically Grounded Subword Segmentation (2024.emnlp-main)
Copied to clipboard
| Challenge: | Statistical word segmentation algorithms have remained a thorn in the side of many researchers. |
| Approach: | They propose to use unsupervised morphological analysis with Morfessor as pre-tokenization and an algebraic method for obtaining subword embeddings grounded in a word embeddable space. |
| Outcome: | The proposed methods improve morphological plausibility and Rényi efficiency on part-of-speech tagging and machine translation tasks. |
Different Time, Different Language: Revisiting the Bias Against Non-Native Speakers in GPT Detectors (2026.eacl-srw)
Copied to clipboard
| Challenge: | Prior work suggests that automated methods falsely flag essays from non-native speakers as generated due to their low perplexity extracted from an LLM, which is supposedly a key feature of the detectors. |
| Approach: | They propose to use a perplexity-based detector to detect essays from non-native speakers of Czech and a detector to examine the effects of different families. |
| Outcome: | The proposed methods are not biased against non-native speakers, but instead falsely flag essays from non-natural speakers as generated, compared to the English essays. |
Combining Static and Contextualised Multilingual Embeddings (2022.findings-acl)
Copied to clipboard
| Challenge: | Static embeddings are less expressive than contextual language models, but can be more straightforwardly aligned across multiple languages. |
| Approach: | They extract static embeddings for 40 languages from XLM-R and validate them with cross-lingual word retrieval and then align them using VecMap. |
| Outcome: | The proposed approach improves multilingual representations by leveraging static embeddings and a pre-training code. |
Speaking Multiple Languages Affects the Moral Bias of Language Models (2023.findings-acl)
Copied to clipboard
Katharina Haemmerl, Bjoern Deiseroth, Patrick Schramowski, Jindřich Libovický, Constantin Rothkopf, Alexander Fraser, Kristian Kersting
| Challenge: | Pre-trained multilingual language models are often better on English than other languages . however, they are trained on varying amounts of data for each language . |
| Approach: | They apply the MORALDIRECTION framework to multilingual models and analyse their results . they find that PMLMs encode differing moral biases, but these do not correspond to cultural differences or commonalities in human opinions. |
| Outcome: | The proposed model captures moral norms from English and imposes them on other languages. |
Is a Prestigious Job the same as a Prestigious Country? A Case Study on Multilingual Sentence Embeddings and European Countries (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Our analysis reveals that the strongest dimension in all models correlates with the political and economic distinction between Western and Eastern Europe and the country’s GDP. |
| Approach: | They propose to prompt models with templated sentences that they machine-translate into 12 European languages and analyze the most prominent dimension in the embeddings. |
| Outcome: | The political distinction between Eastern and Western Europe and the country’s economic strength in terms of GDP are the most prominent features in the embeddings. |
LLM as a Meta-Judge: Synthetic Data for NLP Evaluation Metric Validation (2026.acl-srw)
Copied to clipboard
| Challenge: | Existing evaluation metrics for natural language generation are expensive and time-consuming. |
| Approach: | They propose a framework that utilizes LLMs to generate synthetic evaluation datasets . they propose meta-correlation to measure alignment between metric rankings and human benchmarks based on synthetic data . |
| Outcome: | The proposed framework achieves meta-correlations exceeding 0.9 in multilingual QA and replaces human judgment with synthetic evaluation datasets. |
Multimodal Abstractive Summarization for How2 Videos (P19-1)
Copied to clipboard
| Challenge: | Abstractive summarization is a task of producing a shorter version of the content in the document while preserving its information. |
| Approach: | They propose a new evaluation metric that measures semantic adequacy rather than fluency of abstractive summarization tasks. |
| Outcome: | The proposed model integrates information from different sources into a coherent output. |
How Gender Interacts with Political Values: A Case Study on Czech BERT Models (2024.lrec-main)
Copied to clipboard
| Challenge: | Neural language models are trained on large text corpora that contain value-burdened content and often capture undesirable biases, which the models reflect. |
| Approach: | They propose a method to measure the model's perceived political values by comparing Czech with a representative value survey. |
| Outcome: | The proposed method does not assign statement probability following value-driven reasoning and there is no systematic difference between feminine and masculine sentences. |
Towards Reasonably-Sized Character-Level Transformer NMT by Finetuning Subword Systems (2020.emnlp-main)
Copied to clipboard
| Challenge: | Existing approaches to train character-level models require very deep architectures that are difficult and slow to train. |
| Approach: | They propose to fine tune a Transformer token-based model to get a model without token segmentation. |
| Outcome: | The proposed model improves translation quality and robustness to noise while requiring less token segmentation. |
Why don’t people use character-level machine translation? (2022.findings-acl)
Copied to clipboard
| Challenge: | despite evidence character-level systems are comparable with subword systems, they are rarely used in competitive setups in machine translation competitions. |
| Approach: | They propose a two-step decoder architecture that does not suffer from a slow-down due to the length of character sequences. |
| Outcome: | The proposed character-level MT systems show better domain robustness and better morphological generalization . the proposed decoder architecture shows no slow-down due to the length of character sequences . |
Improving Both Domain Robustness and Domain Adaptability in Machine Translation (2022.coling-1)
Copied to clipboard
| Challenge: | Existing approaches to domain adaptation for NMT depend on high-quality parallel data. |
| Approach: | They propose a meta-learning framework which improves domain robustness and adaptability . they use a word-level domain mixing model and a domain classifier to integrate it . |
| Outcome: | The proposed approach improves domain robustness and adaptability in seen and unseen domains. |
TokCollate: A Comprehensive Tool for Tokenizer Evaluation and Visualization across Languages (2026.acl-demo)
Copied to clipboard
| Challenge: | Tokenization quality varies significantly across languages, leading to disparities in LLM performance and cost for speakers of less-resourced languages. |
| Approach: | They propose a Python-based evaluation framework and a JavaScript visualization interface that evaluates tokenizers in a variety of languages. |
| Outcome: | TokCollate is an evaluation framework for tokenizers with a JavaScript visualization interface. |
End-to-End Non-Autoregressive Neural Machine Translation with Connectionist Temporal Classification (D18-1)
Copied to clipboard
| Challenge: | Autoregressive decoding is the only part of sequence-to-sequence models that prevents massive parallelization at inference time. |
| Approach: | They propose a non-autoregressive architecture based on connectionist temporal classification . they conduct experiments on the WMT English-Romanian and English-German datasets . |
| Outcome: | The proposed model achieves a significant speedup over autoregressive models . the model can be trained end-to-end and maintains translation quality comparable to other models compared to autoregression models based on connectionist temporal classification . |
Probing the Role of Positional Information in Vision-Language Models (2022.findings-naacl)
Copied to clipboard
| Challenge: | Pre-trained Vision-Language models can't be used to understand image structure by injecting position information (PI) about objects in the image. |
| Approach: | They propose two strategies to probe the use of PI in Vision-Language models and investigate their effect on Visual Question Answering. |
| Outcome: | The proposed model can correctly classify if images with detailed PI statements match. |
Charles Translator: A Machine Translation System between Ukrainian and Czech (2024.lrec-main)
Copied to clipboard
Martin Popel, Lucie Polakova, Michal Novák, Jindřich Helcl, Jindřich Libovický, Pavel Straňák, Tomas Krabac, Jaroslava Hlavacova, Mariia Anisimova, Tereza Chlanova
| Challenge: | a system for translating between Ukrainian and Czech was developed in the spring of 2022 . the system was not available at the time in the required quality . |
| Approach: | They propose a machine translation system between Ukrainian and Czech to reduce the impact of the Russian-Ukrainian war on individuals and society. |
| Outcome: | The proposed system translates directly between Ukrainian and Czech, compared to other systems that use English as a pivot. |
On the Language Neutrality of Pre-trained Multilingual Representations (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Existing studies have focused on cross-linguality of contextual embeddings . however, they are only moderately language-neutral by default . |
| Approach: | They propose to use unsupervised centering and fitting an explicit projection on parallel data to achieve stronger language neutrality. |
| Outcome: | The proposed model outperforms existing models on XNLI and NER tasks. |
Exploring Anisotropy and Outliers in Multilingual Language Models for Cross-Lingual Semantic Sentence Similarity (2023.findings-acl)
Copied to clipboard
| Challenge: | Recent studies have shown that contextual language models display outlier dimensions . this is true for monolingual and multilingual models, but little work has been done on multilingual contexts . |
| Approach: | They investigate outlier dimensions and their relationship to anisotropy in multilingual contexts . they focus on cross-lingual semantic similarity tasks . |
| Outcome: | The proposed model improves on cross-lingual semantic similarity tasks. |
Evaluating Morphological Plausibility of Subword Tokenization via Statistical Alignment with Morpho-Syntactic Features (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing metric for subword tokenization evaluation for morphological plausibility requires unavailable or inconsistent gold segmentation data. |
| Approach: | They propose a morpho-syntactic feature-based metric for subword tokenization evaluation. |
| Outcome: | The proposed metric correlates well with traditional morpheme boundary recall while being more broadly applicable across languages with different morphological systems. |
Understanding Cross-Lingual Alignment—A Survey (2024.findings-acl)
Copied to clipboard
| Challenge: | Cross-lingual alignment is the meaningful similarity of representations across languages in multilingual language models. |
| Approach: | They propose a taxonomy of methods to improve cross-lingual alignment . they argue that an effective trade-off between language-neutral and language-specific information is key . |
| Outcome: | The proposed methods can be applied to encoder models and encoder-decoder-only models . they show that language-neutral and language-specific information is key . |