Research
Selected papers, benchmark work, and open research projects.
Selected publications
Lessons Learned from the Evaluation of Portuguese Language Models
Master's thesis introducing Napolab and arguing for more rigorous Portuguese LLM evaluation.
To Tune or Not To Tune? Zero-shot Models for Legal Case Entailment
First-place legal entailment system showing strong zero-shot performance.
Yes, BM25 is a Strong Baseline for Legal Case Retrieval
A useful example of challenging neural-first assumptions with strong baseline analysis.
More papers
Portuguese Language Models and Word Embeddings: Evaluating on Semantic Similarity Tasks
Evaluated Portuguese language models against classical embeddings on semantic similarity benchmarks.
Zero-shot Hashtag Segmentation for Multilingual Sentiment Analysis
Established state-of-the-art results on hashtag segmentation with multilingual transfer.
Earlier distinctions
1st place, ASSIN 2
1st place, COLIEE 2021 Task 2
1st place, ABSAPT 2022