Papers by Leonardo Ranaldi

Multilingual Retrieval-Augmented Generation for Knowledge-Intensive Question Answering Task (2026.findings-eacl)

Challenge:	Recent studies have introduced eclectic strategies to improve reasoning beyond English, but these methods are related to specific language that is not always optimal for reasoning.
Approach:	They propose a modular approach that instructs models to structure reasoning passages in a different problem space and then self-refines their capabilities to deliver step-wise reasoning passage.
Outcome:	The proposed approach achieves significant improvements in multilingual reasoning of various models and task, with improved reasoning consistency across languages.

Copied to clipboard

Leonardo Ranaldi, Barry Haddow, Alexandra Birch

Multilinguality

Challenge:	Existing studies focus on English as the data language for RAG, resulting in limited coverage of multilingual RAG.
Approach:	They propose a method that translates retrieved documents into a common language before generating the response.
Outcome:	The proposed approach improves efficiency on knowledge-intensive tasks but introduces inconsistencies due to cross-lingual variations in the retrieved content.

Empowering cross-lingual abilities of instruction-tuned large language models by translation-following demonstrations (2024.findings-acl)

Copied to clipboard

Leonardo Ranaldi, Giulia Pucci, Andre Freitas

Multilinguality

Challenge:	Large Language Models (LLMs) are often unbalanced towards English because of the imbalance in the distribution of pre-training data.
Approach:	They propose to empower Instruction-tuned LLMs (It-LLMs) in languages other than English by building semantic alignment between them.
Outcome:	The proposed model outperforms the existing model on multilingual question answering benchmarks and adapted versions of MMLU and BBH on six different languages.

Evolving Agents (2026.acl-long)

Copied to clipboard

Leonardo Ranaldi

Challenge:	Current models are static entities incapable of compressing complexity of real world into generalisable concepts . authors: lack of endogenous mechanism for representation updating renders models vulnerable to domain mismatch and catastrophic forgetting .
Approach:	a meta-control system distils on-the-fly abstract representations of states, actions, goals . authors propose a paradigm for autonomous learning driven by pseudo-symbolic abstraction .
Outcome:	a meta-control system distils on-the-fly abstract representations of states, actions, goals . a novel approach resolves the domain mismatch problem and lays the groundwork for truly autonomous AI models .

Eliciting Critical Reasoning in Retrieval-Augmented Generation via Contrastive Explanations (2025.naacl-long)

Copied to clipboard

Leonardo Ranaldi, Marco Valentino, Andre Freitas

R2-MultiOmnia: Leading Multilingual Multimodal Reasoning via Self-Training (2025.acl-long)

Challenge:	Recent studies show that LLMs struggle to critically analyse RAG-based in-context information.
Approach:	They propose a framework that elicits critical arguments in RAG via contrastive explanations . they propose CRAG to retrieve relevant documents given a query and generate explanations that explicitly contrast relevance of passages to support the final answer.
Outcome:	The proposed framework improves state-of-the-art RAG models while requiring significantly fewer prompts and demonstrations and robust to perturbations in the retrieved documents.

Copied to clipboard

Leonardo Ranaldi, Federico Ranaldi, Giulia Pucci

Agentic Oversight via Dialectic Reasoning (2026.acl-long)

Challenge:	Recent studies have introduced eclectic strategies to enhance MLLMs’ reasoning capabilities, but they remain related to a single language.
Approach:	They propose a modular approach that instructs models to abstract key elements of the reasoning process and refine reasoning trajectories via self-correction.
Outcome:	The proposed approach improves multimodal reasoning, gets aligned performances among the languages approaching strong models and improves the model's performance.

Copied to clipboard

Leonardo Ranaldi, Federico Ranaldi

KERMIT: Complementing Transformer Architectures with Encoders of Explicit Syntactic Interpretations (2020.emnlp-main)

Challenge:	Existing approaches to align Large Language Models (LLMs) rely heavily on human annotations, but a Debate between expert models is a promising oversight mechanism.
Approach:	They propose a Debate between expert models to enable scalable oversight . they use a reasoning function to extend the framework to multilingual and multimodal spaces .
Outcome:	The proposed framework outperforms single-expert baselines in six multilingual and multimodal scenarios and shows that argument-mediated supervision instils unsupervised reasoning signals in expert models.

Copied to clipboard

Fabio Massimo Zanzotto, Andrea Santilli, Leonardo Ranaldi, Dario Onorati, Pierfrancesco Tommasino, Francesca Fallucchi

Explainability & Interpretability in NLP

Challenge:	Syntactic parsers are losing their centrality in downstream tasks due to the success of large-scale textual representation learners.
Approach:	They propose to embed symbolic syntactic parse trees into artificial neural networks to visualize how syntax is used in inference.
Outcome:	The proposed encoder can visualize how syntax is used in inference.

Measuring bias in Instruction-Following models with P-AT (2023.findings-emnlp)

Copied to clipboard

Dario Onorati, Elena Ruzzetti, Davide Venditti, Leonardo Ranaldi, Fabio Zanzotto

Improving Multilingual Retrieval-Augmented Language Models through Dialectic Reasoning Argumentations (2025.emnlp-main)

Challenge:	Instruction-Following Language Models (IFLMs) are promising and versatile tools for solving many downstream, information-seeking tasks.
Approach:	They propose a resource to test whether IFLMs are prone to biases . they cast WEAT word tests in promptized classification tasks and associate a metric - the bias score .
Outcome:	The proposed resource consists of 2310 prompts and tests gender and race biases in all the analyzed models.

Copied to clipboard

Leonardo Ranaldi, Federico Ranaldi, Fabio Massimo Zanzotto, Barry Haddow, Alexandra Birch

Aligning Large and Small Language Models via Chain-of-Thought Reasoning (2024.eacl-long)

Challenge:	Existing approaches to ground large language models (LLMs) with RAGs are limited by the heterogeneity of knowledge retrieved.
Approach:	They propose a modular approach guided by Argumentative Explanations that evaluates retrieved information by comparing, contrasting and resolving conflicting perspectives.
Outcome:	The proposed framework significantly improves RAG approaches, requiring low-impact computational effort and providing robustness to knowledge perturbations.

Copied to clipboard

Leonardo Ranaldi, Andre Freitas

Thinking in Schemas: Robust Syllogistic Reasoning in LLMs (2026.acl-long)

Challenge:	Chain-of-Thought (CoT) prompting empowers Large Language Models to solve complex reasoning tasks in a step-wise manner.
Approach:	They propose a method for aligning and transferring reasoning abilities between larger and smaller Language Models by using CoT-Demonstrations.
Outcome:	The proposed method outperforms baselines on question-answering and mathematical reasoning benchmarks.

Copied to clipboard

Federico Ranaldi, Leonardo Ranaldi, Fabio Massimo Zanzotto, Shay B Cohen

Robustness in NLP

Challenge:	syllogistic reasoning models often mistake what sounds true for what is formally valid . content effect is a limitation of sluggish reasoning, which can lead to invalid conclusions . eisape et al., 2024: a key open problem for formal inference in natural language.
Approach:	They propose a schema-guided framework that disentangles semantic plausibility from logical validity.
Outcome:	The proposed framework outperforms existing frameworks while reducing CE.

Learning Multilingual Agentic Policy to Control Sycophancy (2026.eacl-long)

Copied to clipboard

Leonardo Ranaldi, Giulia Pucci

Multilinguality

Challenge:	Large Language Models (LLMs) are effective at adapting to users’ styles, preferences, and contextual signals, but can manifest as sycophancy, i.e., alignment with user-implied beliefs or assumptions even when these contradict factual correctness, uncertainty, or proper logical reasoning.
Approach:	They propose to use large language models to model sycophancy as a decision-making problem by learning agentic policies that are trained to optimise a multi-objective reward that balances task success, scophancies resistance and behavioural consistency.
Outcome:	The proposed model equips a model with an explicit action space that includes answering directly, countering misleading signals, or asking for clarification.

Investigating the Impact of Data Contamination of Large Language Models in Text-to-SQL translation (2024.findings-acl)

Copied to clipboard

Federico Ranaldi, Elena Sofia Ruzzetti, Dario Onorati, Leonardo Ranaldi, Cristina Giannone, Andrea Favalli, Raniero Romagnoli, Fabio Massimo Zanzotto

Text Generation

Challenge:	Large Language Models (LLMs) understand textual description to generate code in zero-shot scenarios, but there is a possibility that this ability may be influenced by having seen target textual descriptions and the related code.
Approach:	They propose a method to detect Data Contamination in Large Language Models (LLMs) and analyze their results on Termite and Spider Datasets to investigate their method.
Outcome:	The proposed method detects data contamination in GPTs and analyzes its performance on unfamiliar datasets.

Advancing Oversight Reasoning across Languages for Audit Sycophantic Behaviour via X-Agent (2025.emnlp-main)

Copied to clipboard

Giulia Pucci, Leonardo Ranaldi

Self-Refine Instruction-Tuning for Aligning Reasoning in Language Models (2024.emnlp-main)

Challenge:	Large language models have demonstrated capabilities that are satisfactory to a wide range of users by adapting to their culture and wisdom.
Approach:	They propose an Oversight Reasoning framework that audits human–LLM dialogues, reasons about them, captures sycophancy and corrects the final outputs.
Outcome:	The proposed framework detects sycophancy, reduces unwarranted agreement and improves cross-turn consistency across different scenarios and languages.

Copied to clipboard

Leonardo Ranaldi, Andre Freitas

Does the Language Matter? Curriculum Learning over Neo-Latin Languages (2024.lrec-main)

Challenge:	Existing approaches to align reasoning abilities between Large Language Models and Smaller Language Model are supervised fine-tuning and preference optimization.
Approach:	They propose a method that elicits Smaller Language Models to self-improve their reasoning abilities via preference optimization.
Outcome:	The proposed method outperforms Instruction-tuning on commonsense and math reasoning tasks on common and math scenarios.

Copied to clipboard

Giulia Pucci, Leonardo Ranaldi

Semantic Text Processing

Challenge:	Curriculum Learning (CL) is emerging as a useful technique to reduce the cost of pre-training Large Language Models.
Approach:	They propose to organize training examples from the simplest to the most complex . they then test the approach to Italian and French to determine the complexity of examples .
Outcome:	The proposed method can be exported to other languages without adaptation.

Lacking the Embedding of a Word? Look it up into a Traditional Dictionary (2022.findings-acl)

Copied to clipboard

Elena Sofia Ruzzetti, Leonardo Ranaldi, Michele Mastromattei, Francesca Fallucchi, Noemi Scarpato, Fabio Massimo Zanzotto

Semantic Text Processing

Challenge:	Word embeddings are powerful dictionaries, but they fail to give sense to rare words . a large body of research is devoted to devising ways to capture word meaning .
Approach:	They propose to use definitions retrieved from traditional dictionaries to build word embeddings for rare words.
Outcome:	The proposed methods outperform state-of-the-art methods for embeddings of unknown words . the proposed methods significantly outperformed the BERT method for OOV words compared to the proposed method .

Dissecting Clinical Reasoning in Natural Language Inference for Large Language Models (2026.findings-acl)

Copied to clipboard

Mael Jullien, Andre Freitas, Marco Valentino, Leonardo Ranaldi

A Tree-of-Thoughts to Broaden Multi-step Reasoning across Languages (2024.findings-naacl)

Challenge:	Recent studies on large language models (LLMs) have demonstrated the impact of prompting strategies and fine-tuning techniques on their reasoning capabilities.
Approach:	They examine four classes of prompting strategies to elicit reasoning in large language models . they then construct demonstrations using a frontier model to distil multi-step reasoning capabilities into smaller models based on Low-Rank Adaptation (LoRA).
Outcome:	The proposed model improves in 75% of the models on MedNLI and TREC Clinical Trials.

Copied to clipboard

Leonardo Ranaldi, Giulia Pucci, Federico Ranaldi, Elena Sofia Ruzzetti, Fabio Massimo Zanzotto

Low-Resource NLP

Challenge:	Existing methods for eliciting Large Language Models (LLMs) to solve complex tasks are limited to English due to the imbalance in the distribution of pre-training data.
Approach:	They propose a method for aligning Cross-lingual CoT reasoning across languages . they propose eliciting Large Language Models to solve complex tasks step-by-step .
Outcome:	The proposed method outperforms existing prompting methods by reducing interactions and achieving state-of-the-art performance.

Can Activation Steering Generalize Across Languages? A Study on Syllogistic Reasoning in Language Models (2026.eacl-long)

Copied to clipboard

Gabriele Maraia, Leonardo Ranaldi, Marco Valentino, Fabio Massimo Zanzotto

Exploring Backward Reasoning in Large Language Models (2025.findings-naacl)

Challenge:	Prior work has focused on activation steering for Large Language Models (LLMs) this technique can be used to improve reasoning accuracy and transferability across languages.
Approach:	They propose to use activation steering to steer models towards a cross-lingual reasoning space.
Outcome:	The proposed techniques generalise well to multilingual datasets while minimizing language modelling performance.

Copied to clipboard

Leonardo Ranaldi, Giulia Pucci

Empowering Multi-step Reasoning across Languages via Program-Aided Language Models (2024.emnlp-main)

Challenge:	Multi-step reasoning through in-context learning strategies have been extensively explored, highlighting the abilities of Large Language Models (LLMs) to solve problems in a step-wise manner.
Approach:	They propose to use Large Language Models to generate answers from step-by-step reasoning by re-constructing the original question that led to the final answer.
Outcome:	The proposed models show that they are able to reason about the conclusion and reconstruct the original question that led to the final answer.

Copied to clipboard

Leonardo Ranaldi, Giulia Pucci, Barry Haddow, Alexandra Birch

When natural language is not enough: The limits of in-context learning demonstrations in multilingual reasoning (2025.findings-naacl)

Challenge:	In-context learning methods elicit Large Language Models to solve tasks using provided demonstrations without parameter updates.
Approach:	They propose a method for aligning reasoning programs across languages using a double-step cross-lingual prompting mechanism.
Outcome:	The proposed method outperforms existing prompting methods and reduces interaction time.

Copied to clipboard

Leonardo Ranaldi, Barry Haddow, Alexandra Birch

HANS, are you clever? Clever Hans Effect Analysis of Neural Systems (2024.starsem-1)

Challenge:	Existing studies have demonstrated the effectiveness of reasoning methods in eliciting multi-step reasoned answers from Large Language Models (LLMs) by leveraging in-context demonstrations.
Approach:	They investigate how well CoT and PAL perform across languages for arithmetic and symbolic reasoning tasks.
Outcome:	The proposed methods perform well in monolingual contexts, primarily in English, but have been limited in other languages.

Copied to clipboard

Leonardo Ranaldi, Fabio Zanzotto

Improving Chain-of-Thought Reasoning via Quasi-Symbolic Abstractions (2025.acl-long)

Challenge:	Large Language Models (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.
Approach:	They propose to use multiple-choice questions (MCQ) benchmarks to assess LLMs' ability to reason around cognitive states, intentions, and reactions of all people involved to investigate their resilience abilities.
Outcome:	The proposed models exhibit exceptional abilities to reason around cognitive states, intentions, and reactions of all people involved, letting humans guide and comprehend day-to-day social interactions effectively.

Copied to clipboard

Leonardo Ranaldi, Marco Valentino, Andre Freitas

A Trip Towards Fairness: Bias and De-Biasing in Large Language Models (2024.starsem-1)

Challenge:	Recent work suggests using logical formalisms coupled with external symbolic solvers to solve complex tasks.
Approach:	They propose a framework to disentangle content from logical reasoning without a complete formalisation.
Outcome:	The proposed methods improve CoT-based methods by up to 8% accuracy on challenging adversarial variations on both natural language and symbolic reasoning tasks.

Copied to clipboard

Leonardo Ranaldi, Elena Ruzzetti, Davide Venditti, Dario Onorati, Fabio Zanzotto

Position Paper: MeMo: Towards Language Models with Associative Memory Mechanisms (2025.findings-acl)

Challenge:	a little or a large bias in CtB-LLMs may cause huge harm . LLaMA and OPT families have an important bias in gender, race, religion, and profession.
Approach:	They propose to debiase three families of Very Large-Language Models with LORA to reduce bias by 4.12 points in the normalized stereotype score.
Outcome:	The proposed model reduces bias up to 4.12 points in the normalized stereotype score.

Copied to clipboard

Fabio Massimo Zanzotto, Elena Sofia Ruzzetti, Giancarlo A. Xompero, Leonardo Ranaldi, Davide Venditti, Federico Ranaldi, Cristina Giannone, Andrea Favalli, Raniero Romagnoli