Arkadiusz Janz

Jan Kocoń, Igor Cichecki, Oliwier Kaszyca et al.

OpenAI has released the Chat Generative Pre-trained Transformer (ChatGPT) and revolutionized the approach in artificial intelligence to human-model interaction. Several publications on ChatGPT evaluation test its effectiveness on well-known natural language processing (NLP) tasks. However, the existing studies are mostly non-automated and tested on a very limited scale. In this work, we examined ChatGPT's capabilities on 25 diverse analytical NLP tasks, most of them subjective even to humans, such as sentiment analysis, emotion recognition, offensiveness, and stance detection. In contrast, the other tasks require more objective reasoning like word sense disambiguation, linguistic acceptability, and question answering. We also evaluated GPT-4 model on five selected subsets of NLP tasks. We automated ChatGPT and GPT-4 prompting process and analyzed more than 49k responses. Our comparison of its results with available State-of-the-Art (SOTA) solutions showed that the average loss in quality of the ChatGPT model was about 25% for zero-shot and few-shot evaluation. For GPT-4 model, a loss for semantic tasks is significantly lower than for ChatGPT. We showed that the more difficult the task (lower SOTA performance), the higher the ChatGPT loss. It especially refers to pragmatic NLP problems like emotion recognition. We also tested the ability to personalize ChatGPT responses for selected subjective tasks via Random Contextual Few-Shot Personalization, and we obtained significantly better user-based predictions. Additional qualitative analysis revealed a ChatGPT bias, most likely due to the rules imposed on human trainers by OpenAI. Our results provide the basis for a fundamental discussion of whether the high quality of recent predictive NLP models can indicate a tool's usefulness to society and how the learning and validation procedures for such systems should be established.

Jan Kocoń, Maciej Piasecki, Arkadiusz Janz et al.

Large Language Models (LLMs) play a central role in modern artificial intelligence, yet their development has been primarily focused on English, resulting in limited support for other languages. We present PLLuM (Polish Large Language Model), the largest open-source family of foundation models tailored specifically for the Polish language. Developed by a consortium of major Polish research institutions, PLLuM addresses the need for high-quality, transparent, and culturally relevant language models beyond the English-centric commercial landscape. We describe the development process, including the construction of a new 140-billion-token Polish text corpus for pre-training, a 77k custom instructions dataset, and a 100k preference optimization dataset. A key component is a Responsible AI framework that incorporates strict data governance and a hybrid module for output correction and safety filtering. We detail the models' architecture, training procedures, and alignment techniques for both base and instruction-tuned variants, and demonstrate their utility in a downstream task within public administration. By releasing these models publicly, PLLuM aims to foster open research and strengthen sovereign AI technologies in Poland.

Łukasz Augustyniak, Kamil Tagowski, Albert Sawczyn et al.

The availability of compute and data to train larger and larger language models increases the demand for robust methods of benchmarking the true progress of LM training. Recent years witnessed significant progress in standardized benchmarking for English. Benchmarks such as GLUE, SuperGLUE, or KILT have become de facto standard tools to compare large language models. Following the trend to replicate GLUE for other languages, the KLEJ benchmark has been released for Polish. In this paper, we evaluate the progress in benchmarking for low-resourced languages. We note that only a handful of languages have such comprehensive benchmarks. We also note the gap in the number of tasks being evaluated by benchmarks for resource-rich English/Chinese and the rest of the world. In this paper, we introduce LEPISZCZE (the Polish word for glew, the Middle English predecessor of glue), a new, comprehensive benchmark for Polish NLP with a large variety of tasks and high-quality operationalization of the benchmark. We design LEPISZCZE with flexibility in mind. Including new models, datasets, and tasks is as simple as possible while still offering data versioning and model tracking. In the first run of the benchmark, we test 13 experiments (task and dataset pairs) based on the five most recent LMs for Polish. We use five datasets from the Polish benchmark and add eight novel datasets. As the paper's main contribution, apart from LEPISZCZE, we provide insights and experiences learned while creating the benchmark for Polish as the blueprint to design similar benchmarks for other low-resourced languages.

Konrad Wojtasik, Vadim Shishkin, Kacper Wołowiec et al.

The BEIR dataset is a large, heterogeneous benchmark for Information Retrieval (IR) in zero-shot settings, garnering considerable attention within the research community. However, BEIR and analogous datasets are predominantly restricted to the English language. Our objective is to establish extensive large-scale resources for IR in the Polish language, thereby advancing the research in this NLP area. In this work, inspired by mMARCO and Mr.~TyDi datasets, we translated all accessible open IR datasets into Polish, and we introduced the BEIR-PL benchmark -- a new benchmark which comprises 13 datasets, facilitating further development, training and evaluation of modern Polish language models for IR tasks. We executed an evaluation and comparison of numerous IR models on the newly introduced BEIR-PL benchmark. Furthermore, we publish pre-trained open IR models for Polish language,d marking a pioneering development in this field. Additionally, the evaluation revealed that BM25 achieved significantly lower scores for Polish than for English, which can be attributed to high inflection and intricate morphological structure of the Polish language. Finally, we trained various re-ranking models to enhance the BM25 retrieval, and we compared their performance to identify their unique characteristic features. To ensure accurate model comparisons, it is necessary to scrutinise individual results rather than to average across the entire benchmark. Thus, we thoroughly analysed the outcomes of IR models in relation to each individual data subset encompassed by the BEIR benchmark. The benchmark data is available at URL {\bf https://huggingface.co/clarin-knext}.

Stanisław Woźniak, Bartłomiej Koptyra, Arkadiusz Janz et al.

Large language models (LLMs) have significantly advanced Natural Language Processing (NLP) tasks in recent years. However, their universal nature poses limitations in scenarios requiring personalized responses, such as recommendation systems and chatbots. This paper investigates methods to personalize LLMs, comparing fine-tuning and zero-shot reasoning approaches on subjective tasks. Results demonstrate that personalized fine-tuning improves model reasoning compared to non-personalized models. Experiments on datasets for emotion recognition and hate speech detection show consistent performance gains with personalized methods across different LLM architectures. These findings underscore the importance of personalization for enhancing LLM capabilities in subjective text perception tasks.

Paweł Walkowiak, Marek Klonowski, Marcin Oleksy et al.

Various techniques are used in the generation of adversarial examples, including methods such as TextBugger which introduce minor, hardly visible perturbations to words leading to changes in model behaviour. Another class of techniques involves substituting words with their synonyms in a way that preserves the text's meaning but alters its predicted class, with TextFooler being a prominent example of such attacks. Most adversarial example generation methods are developed and evaluated primarily on non-inflectional languages, typically English. In this work, we evaluate and explain how adversarial attacks perform in inflectional languages. To explain the impact of inflection on model behaviour and its robustness under attack, we designed a novel protocol inspired by mechanistic interpretability, based on Edge Attribution Patching (EAP) method. The proposed evaluation protocol relies on parallel task-specific corpora that include both inflected and syncretic variants of texts in two languages -- Polish and English. To analyse the models and explain the relationship between inflection and adversarial robustness, we create a new benchmark based on task-oriented dataset MultiEmo, enabling the identification of mechanistic inflection-related elements of circuits within the model and analyse their behaviour under attack.

Piotr Pęzik, Filip Żarnecki, Konrad Kaczyński et al.

This paper describes the instruction dataset used to fine-tune a set of transformer-based large language models (LLMs) developed in the PLLuM (Polish Large Language Model) project. We present a functional typology of the organic, converted, and synthetic instructions used in PLLuM and share some observations about the implications of using human-authored versus synthetic instruction datasets in the linguistic adaptation of base LLMs. Additionally, we release the first representative subset of the PLLuM instruction corpus (PLLuMIC), which we believe to be useful in guiding and planning the development of similar datasets for other LLMs.

Denis Janiak, Julia Moska, Dawid Motyka et al.

Large language models (LLMs) require careful alignment to balance competing objectives - factuality, safety, conciseness, proactivity, and diversity. Existing studies focus on individual techniques or specific dimensions, lacking a holistic assessment of the inherent trade-offs. We propose a unified evaluation framework that compares LLM alignment methods (PPO, DPO, ORPO, KTO) across these five axes, using both in-distribution and out-of-distribution datasets. Leveraging a specialized LLM-as-Judge prompt, validated through human studies, we reveal that DPO and KTO excel in factual accuracy, PPO and DPO lead in safety, and PPO best balances conciseness with proactivity. Our findings provide insights into trade-offs of common alignment methods, guiding the development of more balanced and reliable LLMs.