🧑‍🔬 Leonardo Ranaldi

🏛️ Affiliations

🎓University of Rome Tor VergataRome · Italy2020–2031 · 23 papersEmployment 2023–2024 · Assegnisti, Dipartimento di Ingegneria dell'Impresa "Mario Lucertini"Employment 2025–2031 · Ricercatori a tempo determinato, Dipartimento di Storia, Patrimonio culturale, formazione e societàcurrent🎓University of EdinburghEdinburgh · United Kingdom2023–2026 · 21 papersEmploymentcurrent🔬Idiap Research InstituteMartigny-Combe · Switzerland2023–2026 · 9 papersEmploymentcurrent🎓University of SheffieldSheffield · United Kingdom2026 · 2 paperscurrent🎓Marconi UniversityRome · Italy2020–2023 · 11 papers

From the bylines of their papers (OpenAlex, 2026-10-14), ORCID and Wikidata; institutions on papers are matched by machine. 2 more institutions on single bylines are not listed.

🧭 Fields, topics and concepts

Of its works in our library, with the number of works on each.

📅 Works by year

2020s

📄 Papers

The most cited first (OpenAlex and our library).

1. Aligning Large and Small Language Models via Chain-of-Thought Reasoning (2024 · 70 citations)
   Chain-of-Thought (CoT) prompting helps very large models reason step by step. Smaller models (under about 60B parameters) tend to produce incoherent reasoning when prompted this way, which is a problem for users who can only run compact models. Ranaldi and Freitas propose Instruction-tuning-CoT, a two-stage…
2. Improving Chain-of-Thought Reasoning via Quasi-Symbolic Abstractions (2025 · 42 citations)
   Chain-of-Thought (CoT) prompting helps LLMs reason step by step, but its explanations are prone to content biases that hurt robustness and faithfulness.
3. Self-Refine Instruction-Tuning for Aligning Reasoning in Language Models (2024 · 35 citations)
   Smaller language models (SLMs) usually learn step-wise reasoning through supervised fine-tuning on Chain-of-Thought (CoT) demonstrations from large models (LLMs).
4. SemEval-2026 Task 11: Disentangling Content and Formal Reasoning in Large Language Models (2026 · 22 citations)
   This shared task asks whether large language models can judge if an argument is logically valid without being swayed by whether its content sounds true.
5. Multilingual Retrieval-Augmented Generation for Knowledge-Intensive Task (2025 · 18 citations)
   Most retrieval-augmented generation (RAG) research assumes English questions and English knowledge sources, so it is unclear how well these pipelines work for questions in other languages.
6. Improving Multilingual Retrieval-Augmented Language Models through Dialectic Reasoning Argumentations (2025 · 12 citations)
   Retrieval-augmented generation helps LLMs draw on external facts. In multilingual retrieval, however, the returned documents often disagree, are off-topic, or present different national perspectives, and standard RAG pipelines do not judge their quality. The authors propose Dialectic-RAG (D-RAG), which asks the model…
7. R2-MultiOmnia: Leading Multilingual Multimodal Reasoning via Self-Training (2025 · 6 citations)
   Multimodal LLMs (MLLMs) reason less reliably outside English because high-quality multilingual supervision is scarce.
8. Can Activation Steering Generalize Across Languages? A Study on Syllogistic Reasoning in Language Models (2026 · 5 citations)
   Large language models often judge a syllogism by whether its content sounds believable rather than whether the conclusion actually follows.
9. Learning Multilingual Agentic Policy to Control Sycophancy (2026 · 2 citations)
   The paper addresses sycophancy, where LLMs side with a user's stated or implied beliefs even when those beliefs are factually wrong or poorly reasoned.
10. Sounding vs. Being an Expert: Disentangling Authority, Register and Cultural Impact in Sycophantic LLMs (2026 · 1 citation)
    The paper examines why LLMs abandon correct answers when users push back. It treats sycophancy as a sociolinguistic effect rather than a purely technical flaw. The authors separate two sources of perceived credibility: stated credentials (explicit authority) and polished, formal style (implicit authority through…
11. Agentic Oversight via Dialectic Reasoning (2026)
    The paper asks how weaker models can reliably supervise stronger LLMs. Debate is a known answer, but it offers little verifiable justification for its verdicts and has rarely been tested beyond English text. The authors propose an Agentic Oversight framework built on Dialectic Argumentation. First, they collect an…
12. Thinking in Schemas: Robust Syllogistic Reasoning in LLMs (2026)
    Large language models often accept an argument because its conclusion seems believable, even when the logic does not support it.

🤝 Co-authors here

9 co-authors with a page, the most papers together first.

• Federico Ranaldi ×5
• Andre Freitas ×4
• Fabio Massimo Zanzotto ×4
• Giulia Pucci ×3
• Marco Valentino ×3
• Alexandra Birch ×2
• Barry Haddow ×2
• Gabriele Maraia ×2
• Shay B. Cohen

Venues

Where their works here were published, the most works first, with the editions of those that have a page.

• Annual Meeting of the Association for Computational Linguistics (ACL) ×4: ACL 2026 ×2, ACL 2025 ×2
• Conference of the European Chapter of the Association for Computational Linguistics (EACL) ×3: EACL 2026 ×2, EACL 2024
• Conference on Empirical Methods in Natural Language Processing (EMNLP) ×2: EMNLP 2025, EMNLP 2024
• Findings of the Association for Computational Linguistics ×2: Findings of the Association for Computational Linguistics 2026, Findings of the Association for Computational Linguistics 2025
• International Workshop on Semantic Evaluation (SemEval): SemEval 2026

🔗 Cited by

Papers of our library cite their works 174 times (of the 149,439 papers whose bibliographies we read); 34 citing papers with a page here →.

Updated 1 November 2026

↑ back to top

This page in Markdown → https://md.research.elamur.ai/free/en/authors/leonardo-ranaldi/