AILS-NTUA at SemEval-2025 Task 3: Leveraging Large Language Models and Translation Strategies for Multilingual Hallucination Detection

Multilingual hallucination detection stands as an underexplored challenge, which the Mu-SHROOM shared task seeks to address. In this work, we propose an efficient, training-free LLM prompting strategy that enhances detection by translating multilingual text spans into English. Our approach achieves competitive rankings across multiple languages, securing two first positions in low-resource languages. The consistency of our results highlights the effectiveness of our translation strategy for hallucination detection, demonstrating its applicability regardless of the source language.

AILS-NTUA at SemEval-2025 Task 4: Parameter-Efficient Unlearning for Large Language Models using Data Chunking

The Unlearning Sensitive Content from Large Language Models task aims to remove targeted datapoints from trained models while minimally affecting their general knowledge. In our work, we leverage parameter-efficient, gradient-based unlearning using low-rank (LoRA) adaptation and layer-focused fine-tuning. To further enhance unlearning effectiveness, we employ data chunking, splitting forget data into disjoint partitions and merging them with cyclically sampled retain samples at a pre-defined ratio. Our task-agnostic method achieves an outstanding forget-retain balance, ranking first on leaderboards and significantly outperforming baselines and competing systems.

Pitfalls of Scale: Investigating the Inverse Task of Redefinition in Large Language Models

Inverse tasks can uncover potential reasoning gaps as Large Language Models (LLMs) scale up. In this work, we explore the redefinition task, in which we assign alternative values to well-known physical constants and units of measure, prompting LLMs to respond accordingly. Our findings show that not only does model performance degrade with scale, but its false confidence also rises. Moreover, while factors such as prompting strategies or response formatting are influential, they do not preclude LLMs from anchoring to memorized values.

Bias Beware: The Impact of Cognitive Biases on LLM-Driven Product Recommendations

The advent of Large Language Models (LLMs) has revolutionized product recommendation systems, yet their susceptibility to adversarial manipulation poses critical challenges, particularly in real-world commercial applications. Our approach is the first one to tap into human psychological principles, seamlessly modifying product descriptions, making these adversarial manipulations hard to detect. In this work, we investigate cognitive biases as black-box adversarial strategies, drawing parallels between their effects on LLMs and human purchasing behavior. Through extensive experiments on LLMs of varying scales, we reveal significant vulnerabilities in their use as recommenders, providing critical insights into safeguarding these systems.

BERTtime Stories: Investigating the Role of Synthetic Story Data in Language pre-training

We describe our contribution to the Strict and Strict-Small tracks of the 2nd iteration of the BabyLM Challenge. The shared task is centered around efficient pre-training given data constraints motivated by human development. In response, we study the effect of synthetic story data in language pre-training using TinyStories: a recently introduced dataset of short stories. Initially, we train GPT-Neo models on subsets of TinyStories, while varying the amount of available data. We find that, even with access to less than 100M words, the models are able to generate high-quality, original completions to a given story, and acquire substantial linguistic knowledge. To measure the effect of synthetic story data, we train LTG-BERT encoder models on a combined dataset of: a subset of TinyStories, story completions generated by GPT-Neo, and a subset of the BabyLM dataset. Our experimentation reveals that synthetic data can occasionally offer modest gains, but overall have a negative influence on linguistic understanding. Our work offers an initial study on synthesizing story data in low resource settings and underscores their potential for augmentation in data-constrained language modeling. We publicly release our models and implementation on our GitHub.

RISCORE: Enhancing In-Context Riddle Solving in Language Models through Context-Reconstructed Example Augmentation

Riddle-solving requires advanced reasoning skills, pushing LLMs to engage in abstract thinking and creative problem-solving, often revealing limitations in their cognitive abilities. In this paper, we examine the riddle-solving capabilities of LLMs using a multiple-choice format, exploring how different prompting techniques impact performance on riddles that demand diverse reasoning skills. To enhance results, we introduce RISCORE (RIddle Solving with COntext REcontruciton) a novel fully automated prompting method that generates and utilizes contextually reconstructed sentence-based puzzles in conjunction with the original examples to create few-shot exemplars. Our experiments demonstrate that RISCORE significantly improves the performance of language models in both vertical and lateral thinking tasks, surpassing traditional exemplar selection strategies across a variety of few-shot settings.

"I Never Said That": A dataset, taxonomy and baselines on response clarity classification

Equivocation and ambiguity in public speech are well-studied discourse phenomena, especially in political science and analysis of political interviews. Inspired by the well-grounded theory on equivocation, we aim to resolve the closely related problem of response clarity in questions extracted from political interviews, leveraging the capabilities of Large Language Models (LLMs) and human expertise. To this end, we introduce a novel taxonomy that frames the task of detecting and classifying response clarity and a corresponding clarity classification dataset which consists of question-answer (QA) pairs drawn from political interviews and annotated accordingly. Our proposed two-level taxonomy addresses the clarity of a response in terms of the information provided for a given question (high-level) and also provides a fine-grained taxonomy of evasion techniques that relate to unclear, ambiguous responses (lower-level). We combine ChatGPT and human annotators to collect, validate and annotate discrete QA pairs from political interviews, to be used for our newly introduced response clarity task. We provide a detailed analysis and conduct several experiments with different model architectures, sizes and adaptation methods to gain insights and establish new baselines over the proposed dataset and task.

Enhancing adversarial robustness in Natural Language Inference using explanations

The surge of state-of-the-art Transformer-based models has undoubtedly pushed the limits of NLP model performance, excelling in a variety of tasks. We cast the spotlight on the underexplored task of Natural Language Inference (NLI), since models trained on popular well-suited datasets are susceptible to adversarial attacks, allowing subtle input interventions to mislead the model. In this work, we validate the usage of natural language explanation as a model-agnostic defence strategy through extensive experimentation: only by fine-tuning a classifier on the explanation rather than premise-hypothesis inputs, robustness under various adversarial attacks is achieved in comparison to explanation-free baselines. Moreover, since there is no standard strategy of testing the semantic validity of the generated explanations, we research the correlation of widely used language generation metrics with human perception, in order for them to serve as a proxy towards robust NLI models. Our approach is resource-efficient and reproducible without significant computational limitations.

Prompt2Fashion: An automatically generated fashion dataset

Despite the rapid evolution and increasing efficacy of language and vision generative models, there remains a lack of comprehensive datasets that bridge the gap between personalized fashion needs and AI-driven design, limiting the potential for truly inclusive and customized fashion solutions. In this work, we leverage generative models to automatically construct a fashion image dataset tailored to various occasions, styles, and body types as instructed by users. We use different Large Language Models (LLMs) and prompting strategies to offer personalized outfits of high aesthetic quality, detail, and relevance to both expert and non-expert users' requirements, as demonstrated by qualitative analysis. Up until now the evaluation of the generated outfits has been conducted by non-expert human subjects. Despite the provided fine-grained insights on the quality and relevance of generation, we extend the discussion on the importance of expert knowledge for the evaluation of artistic AI-generated datasets such as this one. Our dataset is publicly available on GitHub at https://github.com/georgiarg/Prompt2Fashion.

ARPA: A Novel Hybrid Model for Advancing Visual Word Disambiguation Using Large Language Models and Transformers

In the rapidly evolving fields of natural language processing and computer vision, Visual Word Sense Disambiguation (VWSD) stands as a critical, yet challenging task. The quest for models that can seamlessly integrate and interpret multimodal data is more pressing than ever. Imagine a system that can understand language with the depth and nuance of human cognition, while simultaneously interpreting the rich visual context of the world around it. We present ARPA, an architecture that fuses the unparalleled contextual understanding of large language models with the advanced feature extraction capabilities of transformers, which then pass through a custom Graph Neural Network (GNN) layer to learn intricate relationships and subtle nuances within the data. This innovative architecture not only sets a new benchmark in visual word disambiguation but also introduces a versatile framework poised to transform how linguistic and visual data interact by harnessing the synergistic strengths of its components, ensuring robust performance even in the most complex disambiguation scenarios. Through a series of experiments and comparative analysis, we reveal the substantial advantages of our model, underscoring its potential to redefine standards in the field. Beyond its architectural prowess, our architecture excels through experimental enrichments, including sophisticated data augmentation and multi-modal training techniques. ARPA's introduction marks a significant milestone in visual word disambiguation, offering a compelling solution that bridges the gap between linguistic and visual modalities. We invite researchers and practitioners to explore the capabilities of our model, envisioning a future where such hybrid models drive unprecedented advancements in artificial intelligence.