Vitor Gaboardi dos Santos
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Vitor Gaboardi dos Santos has a total of 35 co-authors. Here are the top co-authors by number of shared publications:
- Theo Lynn
Dublin City University
5 shared publications - Boualem Benatallah
Dublin City University
4 shared publications - Guto Leoni Santos
Dublin City University
4 shared publications - Patricia Takako Endo
Programa de Pós-Graduação em Engenharia da Computação (PPGEC), Universidade de Pernambuco (UPE), Recife, Brazil
4 shared publications - Colm Kearns
Dublin City University
3 shared publications - Daniel Kilvington
Leeds Beckett University, UK
3 shared publications
Latest Publications
- 13 October 2026
- Published by Elsevier in Information and Software Technology
- Vol. 201
DAP: A diversity-aware pipeline for LLM-based utterance generation and API retrieval
- 13 October 2026
- Published by Elsevier in Information and Software Technology
- Vol. 201
DAP: A diversity-aware pipeline for LLM-based utterance generation and API retrieval
- 7 June 2026
- Published by Springer Nature
- p. 20-27
Constraint-Aware Pipeline for Generating High-Quality Utterances from OpenAPI Specifications Using LLMs
- 7 June 2026
- Published by Springer Nature
- p. 20-27
Constraint-Aware Pipeline for Generating High-Quality Utterances from OpenAPI Specifications Using LLMs
- Open Access
- 25 February 2026
- Published by Springer Nature in Journal of Big Data
- Vol. 13 (1)
All together against hate: ensemble based LLMs for multi class hate speech classification in the football context
- 25 February 2026
- Published by Springer Nature in Journal of Big Data
- Vol. 13 (1)
Open AccessAll together against hate: ensemble based LLMs for multi class hate speech classification in the football context
- Open Access
- 3 February 2026
- Published by Elsevier in SSRN Electronic Journal
DAP: A Diversity-Aware Pipeline for LLM-Based Utterance Generation and API Retrieval
Context: The integration of Large Language Models (LLMs) with APIs enables access to real-time information and supports the execution of actions in digital and physical software systems. However, challenges in this integration include handling the linguistic diversity of user utterances and accurately retrieving relevant API methods given the plethora of available APIs.Objectives: This study introduces DAP, an AI-driven middleware pipeline that leverages LLMs to systematically generate linguistically diverse utterances from OpenAPI Specifications (OAS) and to build robust representations for API method retrieval.Methods: DAP consists of three main stages: (1) generating initial utterances using LLMs and data provided by OAS; (2) filtering errors and paraphrasing unique utterances to enrich lexical, syntactic, and semantic diversity using LLMs; and (3) building API method vectors using embedding models based on the LLM-generated diversity-aware utterances. Results: DAP is evaluated using existing benchmark datasets, comparing utterance generation across five LLMs and API method vector representation across three embedding models. The results show that DAP consistently increases linguistic diversity relative to a baseline utterance generation pipeline and using diversity-aware paraphrases improves API method retrieval performance by up to 10\% in NDCG@10 compared to documentation-only representations.Conclusion: The results demonstrate that diversity-aware, LLM-based utterance generation can serve as an effective mechanism for improving API method retrieval. More broadly, the findings suggest that alternative, language-centric representations of API methods can complement or surpass traditional technical documentation in AI-driven software systems.
- 3 February 2026
- Published by Elsevier in SSRN Electronic Journal
Open AccessDAP: A Diversity-Aware Pipeline for LLM-Based Utterance Generation and API Retrieval
Context: The integration of Large Language Models (LLMs) with APIs enables access to real-time information and supports the execution of actions in digital and physical software systems. However, challenges in this integration include handling the linguistic diversity of user utterances and accurately retrieving relevant API methods given the plethora of available APIs.Objectives: This study introduces DAP, an AI-driven middleware pipeline that leverages LLMs to systematically generate linguistically diverse utterances from OpenAPI Specifications (OAS) and to build robust representations for API method retrieval.Methods: DAP consists of three main stages: (1) generating initial utterances using LLMs and data provided by OAS; (2) filtering errors and paraphrasing unique utterances to enrich lexical, syntactic, and semantic diversity using LLMs; and (3) building API method vectors using embedding models based on the LLM-generated diversity-aware utterances. Results: DAP is evaluated using existing benchmark datasets, comparing utterance generation across five LLMs and API method vector representation across three embedding models. The results show that DAP consistently increases linguistic diversity relative to a baseline utterance generation pipeline and using diversity-aware paraphrases improves API method retrieval performance by up to 10\% in NDCG@10 compared to documentation-only representations.Conclusion: The results demonstrate that diversity-aware, LLM-based utterance generation can serve as an effective mechanism for improving API method retrieval. More broadly, the findings suggest that alternative, language-centric representations of API methods can complement or surpass traditional technical documentation in AI-driven software systems.
- Open Access
ALRM: Agentic LLM for Robotic Manipulation
Large Language Models (LLMs) have recently empowered agentic frameworks to exhibit advanced reasoning and planning capabilities. However, their integration in robotic control pipelines remains limited in two aspects: (1) prior \ac{llm}-based approaches often lack modular, agentic execution mechanisms, limiting their ability to plan, reflect on outcomes, and revise actions in a closed-loop manner; and (2) existing benchmarks for manipulation tasks focus on low-level control and do not systematically evaluate multistep reasoning and linguistic variation. In this paper, we propose Agentic LLM for Robot Manipulation (ALRM), an LLM-driven agentic framework for robotic manipulation. ALRM integrates policy generation with agentic execution through a ReAct-style reasoning loop, supporting two complementary modes: Code-asPolicy (CaP) for direct executable control code generation, and Tool-as-Policy (TaP) for iterative planning and tool-based action execution. To enable systematic evaluation, we also introduce a novel simulation benchmark comprising 56 tasks across multiple environments, capturing linguistically diverse instructions. Experiments with ten LLMs demonstrate that ALRM provides a scalable, interpretable, and modular approach for bridging natural language reasoning with reliable robotic execution. Results reveal Claude-4.1-Opus as the top closed-source model and Falcon-H1-7B as the top open-source model under CaP.
Open AccessALRM: Agentic LLM for Robotic Manipulation
Large Language Models (LLMs) have recently empowered agentic frameworks to exhibit advanced reasoning and planning capabilities. However, their integration in robotic control pipelines remains limited in two aspects: (1) prior \ac{llm}-based approaches often lack modular, agentic execution mechanisms, limiting their ability to plan, reflect on outcomes, and revise actions in a closed-loop manner; and (2) existing benchmarks for manipulation tasks focus on low-level control and do not systematically evaluate multistep reasoning and linguistic variation. In this paper, we propose Agentic LLM for Robot Manipulation (ALRM), an LLM-driven agentic framework for robotic manipulation. ALRM integrates policy generation with agentic execution through a ReAct-style reasoning loop, supporting two complementary modes: Code-asPolicy (CaP) for direct executable control code generation, and Tool-as-Policy (TaP) for iterative planning and tool-based action execution. To enable systematic evaluation, we also introduce a novel simulation benchmark comprising 56 tasks across multiple environments, capturing linguistically diverse instructions. Experiments with ten LLMs demonstrate that ALRM provides a scalable, interpretable, and modular approach for bridging natural language reasoning with reliable robotic execution. Results reveal Claude-4.1-Opus as the top closed-source model and Falcon-H1-7B as the top open-source model under CaP.
- 15 June 2025
- Published by Springer Nature
- p. 145-162
LLMs to Replace Crowdsourcing in Generating Syntactically Diverse Paraphrases for Task-Oriented Chatbots
- 15 June 2025
- Published by Springer Nature
- p. 145-162
LLMs to Replace Crowdsourcing in Generating Syntactically Diverse Paraphrases for Task-Oriented Chatbots
- 24 January 2025
- Published by Springer Nature
- p. 489-504
Detecting Homophobic Speech in Soccer Tweets Using Large Language Models and Explainable AI
- 24 January 2025
- Published by Springer Nature
- p. 489-504
Detecting Homophobic Speech in Soccer Tweets Using Large Language Models and Explainable AI
- 3 June 2024
- Published by Springer Nature
- p. 531-546
Identifying Citizen-Related Issues from Social Media Using LLM-Based Data Augmentation
- 3 June 2024
- Published by Springer Nature
- p. 531-546
Identifying Citizen-Related Issues from Social Media Using LLM-Based Data Augmentation
- 3 June 2024
- Published by Springer Nature
- p. 547-562
Kicking Prejudice: Large Language Models for Racism Classification in Soccer Discourse on Social Media
- 3 June 2024
- Published by Springer Nature
- p. 547-562
Kicking Prejudice: Large Language Models for Racism Classification in Soccer Discourse on Social Media