Search

The cognitive essence of humans is deeply intertwined with the concept of animacy, which plays an essential role in shaping their memory, vision, and multi-layered language understanding. Although animacy appears in language via nuanced constraints on verbs and adjectives, it is also learned and refined through extralinguistic information. Similarly, we assume that the LLMs' limited abilities to understand natural language when processing animacy are motivated by the fact that these models are trained exclusively on text. Hence, the question this paper aims to answer arises: can LLMs, in their digital wisdom, process animacy in a similar way to what humans would do? We then propose a systematic analysis via prompting approaches. In particular, we probe different LLMs by prompting them using animate, inanimate, usual, and stranger contexts. Results reveal that, although LLMs have been trained predominantly on textual data, they exhibit human-like behavior when faced with typical animate and inanimate entities in alignment with earlier studies. Hence, LLMs can adapt to understand unconventional situations by recognizing oddities as animated without needing to interface with unspoken cognitive triggers humans rely on to break down animations.

Large Language Models (LLMs) represent a significant advancement in artificial intelligence, finding applications across various domains. However, their reliance on massive internet-sourced datasets for training brings notable privacy issues, which are exacerbated in critical domains (e.g., healthcare). Moreover, certain application-specific scenarios may require fine-tuning these models on private data. This survey critically examines the privacy threats associated with LLMs, emphasizing the potential for these models to memorize and inadvertently reveal sensitive information. We explore current threats by reviewing privacy attacks on LLMs and propose comprehensive solutions for integrating privacy mechanisms throughout the entire learning pipeline. These solutions range from anonymizing training datasets to implementing differential privacy during training or inference and machine unlearning after training. Our comprehensive review of existing literature highlights ongoing challenges, available tools, and future directions for preserving privacy in LLMs. This work aims to guide the development of more secure and trustworthy AI systems by providing a thorough understanding of privacy preservation methods and their effectiveness in mitigating risks., Comment: GitHub repository: https://github.com/michele17284/Awesome-Privacy-Preserving-LLMs

Large language models (LLMs) require a significant redesign in solutions to preserve privacy in data-intensive applications due to their text-generation capabilities. Indeed, LLMs tend to memorize and emit private information when maliciously prompted. In this paper, we introduce Private Association Editing (PAE) as a novel defense approach for private data leakage. PAE is designed to effectively remove Personally Identifiable Information (PII) without retraining the model. Experimental results demonstrate the effectiveness of PAE with respect to alternative baseline methods. We believe PAE will serve as a critical tool in the ongoing effort to protect data privacy in LLMs, encouraging the development of safer models for real-world applications.

This paper explores the correlation between linguistic diversity, sentiment analysis and transformer model architectures. We aim to investigate how different English variations impact transformer-based models for irony detection. To conduct our study, we used the EPIC corpus to extract five diverse English variation-specific datasets and applied the KEN pruning algorithm on five different architectures. Our results reveal several similarities between optimal subnetworks, which provide insights into the linguistic variations that share strong resemblances and those that exhibit greater dissimilarities. We discovered that optimal subnetworks across models share at least 60% of their parameters, emphasizing the significance of parameter values in capturing and interpreting linguistic variations. This study highlights the inherent structural similarities between models trained on different variants of the same language and also the critical role of parameter values in capturing these nuances.

Understanding textual description to generate code seems to be an achieved capability of instruction-following Large Language Models (LLMs) in zero-shot scenario. However, there is a severe possibility that this translation ability may be influenced by having seen target textual descriptions and the related code. This effect is known as Data Contamination. In this study, we investigate the impact of Data Contamination on the performance of GPT-3.5 in the Text-to-SQL code-generating tasks. Hence, we introduce a novel method to detect Data Contamination in GPTs and examine GPT-3.5's Text-to-SQL performances using the known Spider Dataset and our new unfamiliar dataset Termite. Furthermore, we analyze GPT-3.5's efficacy on databases with modified information via an adversarial table disconnection (ATD) approach, complicating Text-to-SQL tasks by removing structural pieces of information from the database. Our results indicate a significant performance drop in GPT-3.5 on the unfamiliar Termite dataset, even with ATD modifications, highlighting the effect of Data Contamination on LLMs in Text-to-SQL translation tasks.

Neural network pruning has become increasingly crucial due to the complexity of these models and their widespread use in various fields. Existing pruning algorithms often suffer from limitations such as architecture specificity, excessive complexity and reliance on demanding calculations, rendering them impractical for real-world applications. This paper introduces KEN: a straightforward, universal and unstructured pruning algorithm based on Kernel Density Estimation (KDE). KEN aims to construct optimized transformers by selectively preserving the most significant parameters while restoring others to their pre-training state. This strategy preserves model performance while enabling storage of only the optimized subnetwork, leading to substantial memory savings. Extensive evaluations across seven different LLMs demonstrate that KEN achieves equal or better performance than their original unpruned versions, with a minimum parameter reduction of 25%. Furthermore, in-depth comparisons with established pruning and PEFT algorithms confirm KEN effectiveness. We further introduce KEN$_{viz}$, an explainable tool that visualizes the optimized model composition achieved by KEN from different points of view., Comment: 14 pages, 8 figures, 7 tables - to be published at ACL 2024

Reasoning methods, best exemplified by the well-known Chain-of-Thought (CoT), empower the reasoning abilities of Large Language Models (LLMs) by eliciting them to solve complex tasks in a step-by-step manner. Although they are achieving significant success, the ability to deliver multi-step reasoning remains limited to English because of the imbalance in the distribution of pre-training data, which makes other languages a barrier. In this paper, we propose Cross-lingual Tree-of-Thoughts (Cross-ToT), a method for aligning Cross-lingual CoT reasoning across languages. The proposed method, through a self-consistent cross-lingual prompting mechanism inspired by the Tree-of-Thoughts approach, provides multi-step reasoning paths in different languages that, during the steps, lead to the final solution. Experimental evaluations show that our method significantly outperforms existing prompting methods by reducing the number of interactions and achieving state-of-the-art performance., Comment: Findings of the Association for Computational Linguistics: NAACL 2024

Instruction-tuned Large Language Models (It-LLMs) have been exhibiting outstanding abilities to reason around cognitive states, intentions, and reactions of all people involved, letting humans guide and comprehend day-to-day social interactions effectively. In fact, several multiple-choice questions (MCQ) benchmarks have been proposed to construct solid assessments of the models' abilities. However, earlier works are demonstrating the presence of inherent "order bias" in It-LLMs, posing challenges to the appropriate evaluation. In this paper, we investigate It-LLMs' resilience abilities towards a series of probing tests using four MCQ benchmarks. Introducing adversarial examples, we show a significant performance gap, mainly when varying the order of the choices, which reveals a selection bias and brings into discussion reasoning abilities. Following a correlation between first positions and model choices due to positional bias, we hypothesized the presence of structural heuristics in the decision-making process of the It-LLMs, strengthened by including significant examples in few-shot scenarios. Finally, by using the Chain-of-Thought (CoT) technique, we elicit the model to reason and mitigate the bias by obtaining more robust models., Comment: This paper contains erroneous evaluations and we would like to withdraw it

Cheap-to-Build Very Large-Language Models (CtB-LLMs) with affordable training are emerging as the next big revolution in natural language processing and understanding. These CtB-LLMs are democratizing access to trainable Very Large-Language Models (VLLMs) and, thus, may represent the building blocks of many NLP systems solving downstream tasks. Hence, a little or a large bias in CtB-LLMs may cause huge harm. In this paper, we performed a large investigation of the bias of three families of CtB-LLMs, and we showed that debiasing techniques are effective and usable. Indeed, according to current tests, the LLaMA and the OPT families have an important bias in gender, race, religion, and profession. In contrast to the analysis for other LLMs, we discovered that bias depends not on the number of parameters but on the perplexity. Finally, the debiasing of OPT using LoRA reduces bias up to 4.12 points in the normalized stereotype score.

Pre-trained Language Models such as BERT are impressive machines with the ability to memorize, possibly generalized learning examples. We present here a small, focused contribution to the analysis of the interplay between memorization and performance of BERT in downstream tasks. We propose PreCog, a measure for evaluating memorization from pre-training, and we analyze its correlation with the BERT's performance. Our experiments show that highly memorized examples are better classified, suggesting memorization is an essential key to success for BERT.

The impressive achievements of transformers force NLP researchers to delve into how these models represent the underlying structure of natural language. In this paper, we propose a novel standpoint to investigate the above issue: using typological similarities among languages to observe how their respective monolingual models encode structural information. We aim to layer-wise compare transformers for typologically similar languages to observe whether these similarities emerge for particular layers. For this investigation, we propose to use Centered Kernel Alignment to measure similarity among weight matrices. We found that syntactic typological similarity is consistent with the similarity between the weights in the middle layers, which are the pretrained BERT layers to which syntax encoding is generally attributed. Moreover, we observe that a domain adaptation on semantically equivalent texts enhances this similarity among weight matrices.

By using modular functions on the upper complex half-plane, we study a class of strain energies for crystalline materials whose global invariance originates from the full symmetry group of the underlying lattice. This follows Ericksen's suggestion which aimed at extending the Landau-type theories to encompass the behavior of crystals undergoing structural phase transformation, with twinning, microstructure formation, and possibly associated plasticity effects. Here we investigate such Ericksen-Landau strain energies for the modelling of reconstructive transformations, focusing on the prototypical case of the square-hexagonal phase change in 2D crystals. We study the bifurcation and valley-floor network of these potentials, and use one in the simulation of a quasi-static shearing test. We observe typical effects associated with the micro-mechanics of phase transformation in crystals, in particular, the bursty progression of the structural phase change, characterized by intermittent stress-relaxation through microstructure formation, mediated, in this reconstructive case, by defect nucleation and movement in the lattice., Comment: 17 pages, 6 figures, links to 4 supplementary videos

Machine Learning may push research in precision medicine to unprecedented heights. To succeed, machine learning needs a large amount of data, often including personal data. Therefore, machine learning applied to precision medicine is on a cliff edge: if it does not learn to fly, it will deeply fall down. In this paper, we present Active Informed Consent (AIC) as a novel hybrid legal-technological tool to foster the gathering of a large amount of data for machine learning. We carefully analyzed the compliance of this technological tool to the legal intricacies protecting the privacy of European Citizens.

Pre-trained Transformers are challenging human performances in many NLP tasks. The massive datasets used for pre-training seem to be the key to their success on existing tasks. In this paper, we explore how a range of pre-trained Natural Language Understanding models perform on definitely unseen sentences provided by classification tasks over a DarkNet corpus. Surprisingly, results show that syntactic and lexical neural networks perform on par with pre-trained Transformers even after fine-tuning. Only after what we call extreme domain adaptation, that is, retraining with the masked language model task on all the novel corpus, pre-trained Transformers reach their standard high results. This suggests that huge pre-training corpora may give Transformers unexpected help since they are exposed to many of the possible sentences.

Incorporating explicit domain knowledge into neural-based task-oriented dialogue systems is an effective way to reduce the need of large sets of annotated dialogues. In this paper, we investigate how the use of explicit domain knowledge of conversational designers affects the performance of neural-based dialogue systems. To support this investigation, we propose the Conversational-Logic-Injection-in-Neural-Network system (CLINN) where explicit knowledge is coded in semi-logical rules. By using CLINN, we evaluated semi-logical rules produced by a team of differently skilled conversational designers. We experimented with the Restaurant topic of the MultiWOZ dataset. Results show that external knowledge is extremely important for reducing the need of annotated examples for conversational systems. In fact, rules from conversational designers used in CLINN significantly outperform a state-of-the-art neural-based dialogue system.

Word embeddings are powerful dictionaries, which may easily capture language variations. However, these dictionaries fail to give sense to rare words, which are surprisingly often covered by traditional dictionaries. In this paper, we propose to use definitions retrieved in traditional dictionaries to produce word embeddings for rare words. For this purpose, we introduce two methods: Definition Neural Network (DefiNNet) and Define BERT (DefBERT). In our experiments, DefiNNet and DefBERT significantly outperform state-of-the-art as well as baseline methods devised for producing embeddings of unknown words. In fact, DefiNNet significantly outperforms FastText, which implements a method for the same task-based on n-grams, and DefBERT significantly outperforms the BERT method for OOV words. Then, definitions in traditional dictionaries are useful to build word embeddings for rare words.

We study the mechanical response of a dislocation-free 2D crystal under homogenous shear using a new mesoscopic approach to crystal plasticity, a Landau-type theory, accounting for the global invariance of the energy in the space of strain tensors while operating with an infinite number of equivalent energy wells. The advantage of this approach is that it eliminates arbitrariness in dealing with topological transitions involved, for instance, in nucleation and annihilation of dislocations. We use discontinuous yielding of pristine micro-crystals as a benchmark problem for the new theory and show that the nature of the catastrophic instability, which in this setting inevitably follows the standard affine response, depends not only on lattice symmetry but also on the orientation of the crystal in the loading device. The ensuing dislocation avalanche involves cooperative dislocation nucleation, resulting in the formation of complex microstructures controlled by a nontrivial self-induced coupling between different plastic mechanisms.

Artificial Intelligence is providing astonishing results, with medicine being one of its favourite playgrounds. In a few decades, computers may be capable of formulating diagnoses and choosing the correct treatment, while robots may perform surgical operations, and conversational agents could interact with patients as virtual coaches. Machine Learning and, in particular, Deep Neural Networks are behind this revolution. In this scenario, important decisions will be controlled by standalone machines that have learned predictive models from provided data. Among the most challenging targets of interest in medicine are cancer diagnosis and therapies but, to start this revolution, software tools need to be adapted to cover the new requirements. In this sense, learning tools are becoming a commodity in Python and Matlab libraries, just to name two, but to exploit all their possibilities, it is essential to fully understand how models are interpreted and which models are more interpretable than others. In this survey, we analyse current machine learning models, frameworks, databases and other related tools as applied to medicine - specifically, to cancer research - and we discuss their interpretability, performance and the necessary input data. From the evidence available, ANN, LR and SVM have been observed to be the preferred models. Besides, CNNs, supported by the rapid development of GPUs and tensor-oriented programming libraries, are gaining in importance. However, the interpretability of results by doctors is rarely considered which is a factor that needs to be improved. We therefore consider this study to be a timely contribution to the issue., Comment: 64 pages, 3 figures, 7 tables, +100 references

Creativity is one of the driving forces of human kind as it allows to break current understanding to envision new ideas, which may revolutionize entire fields of knowledge. Scientific research offers a challenging environment where to learn a model for the creative process. In fact, scientific research is a creative act in the formal settings of the scientific method and this creative act is described in articles. In this paper, we dare to introduce the novel, scientifically and philosophically challenging task of Generating Abstracts of Scientific Papers from abstracts of cited papers (GASP) as a text-to-text task to investigate scientific creativity, To foster research in this novel, challenging task, we prepared a dataset by using services where that solve the problem of copyright and, hence, the dataset is public available with its standard split. Finally, we experimented with two vanilla summarization systems to start the analysis of the complexity of the GASP task., Comment: 8 pages

We explore the nonlinear variational modelling of two-dimensional (2D) crystal plasticity based on strain energies which are invariant under the full symmetry group of 2D lattices. We use a natural parameterization of strain space via the upper complex Poincar\'e half-plane. This transparently displays the constraints imposed by lattice symmetry on the energy landscape. Quasi-static energy minimization naturally induces bursty plastic flow and shape change in the crystal due to the underlying coordinated basin-hopping local strain activity. This is mediated by the nucleation, interaction, and annihilation of lattice defects occurring with no need for auxiliary hypotheses. Numerical simulations highlight the marked effect of symmetry on all these processes. The kinematical atlas induced by symmetry on strain space elucidates how the arrangement of the energy extremals and the possible bifurcations of the strain-jump paths affect the plastification mechanisms and defect-pattern complexity in the lattice., Comment: 24 pages, 4 figures

We show that nonlinear continuum elasticity can be effective in modeling plastic flows in crystals if it is viewed as Landau theory with an infinite number of equivalent energy wells whose configuration is dictated by the symmetry group GL(3,Z). Quasi-static loading can be then handled by athermal dynamics, while lattice based discretization can play the role of regularization. As a proof of principle, we study in this Letter dislocation nucleation in a homogeneously sheared 2D crystal and show that the global tensorial invariance of the elastic energy foments the development of complexity in the configuration of collectively nucleating defects. A crucial role in this process is played by the unstable higher symmetry crystallographic phases, traditionally thought to be unrelated to plastic flow in lower symmetry lattices.