Latest Dispatch

FIELD NOTE — I've been experimenting with a live leaderboard system—integrating GitHub Classroom with local CI/CD runners—hoping to make the labs more interactive. Watching 29 student teams engage with their code in real-time has been a deeply rewarding teaching experience.

FIELD NOTE — It’s one thing to have a model working on a server, and another to actually let people play with it. I finally put together a demo for REFINE-LM—a small step towards making our research on fairness and bias reduction a bit more tangible for everyone.

FIELD NOTE — Exciting week in Santiago de Compostela! I presented our latest research 'REFINE-LM: Mitigating Language Model Stereotypes via Reinforcement Learning' at the European Conference on Artificial Intelligence. We discussed novel fine-tuning methods that leverage rewards based on fairness metrics.

FIELD NOTE — New proceedings are out! Our paper 'The Balancing Act' was presented at the Fourth Workshop on LT-EDI. We explored the subtle biases in Automated Speech Recognition (ASR) specifically for the Portuguese language, proposing methods to unmask and alleviate these disparities.

FIELD NOTE — Honored to be part of the findings for the First Workshop on Simulating Conversational Intelligence in Chat (SCI-CHAT). As conversational AI continues to evolve, our findings at EACL 2024 highlight critical benchmarks for human-like intelligence in digital assistants.

FIELD NOTE — Published a new preprint on HAL. We've been working on a reinforcement learning approach to mitigate stereotypical biases in Language Models, exploring how adaptive reward systems can steer models toward more neutral outputs.

FIELD NOTE — Successfully shared my research on 'Mitigating Unintended Bias in Masked Language Models'. This work laid the foundation for our current investigations into reinforcement learning and fairness in NLP.

FIELD NOTE — I've been wrestling with the concept of 'meaning' lately—a mystery that feels far too big for a single series. What started as simple curiosity turned into a two-part reflection on linguistics and evolution. It’s mostly just an attempt to organize my own thoughts on how we connect our minds to the world.

FIELD NOTE — Launched NewsChase, a mobile application designed to fight manipulative content in digital journalism. By assigning a 'manipulative index' to articles, we aim to educate readers and foster a more cautious approach to news consumption.

FIELD NOTE — Presented 'A Simple Approach to Classify Fictional and Non-Fictional Genres' at the Storytelling Workshop. It was an incredible experience meeting the global NLP community in Florence.

FIELD NOTE — Started working on a small tool to see if machines can actually 'feel' the difference between a novel and a news report. It's been a lot of trial and error with linguistic features, but it’s been fascinating to see how the code starts to recognize the rhythm of storytelling.