Nuno Fachada

Can Large Language Models Implement Agent-Based Models? An ODD-based Replication Study Large language models (LLMs) can now synthesize non-trivial executable code from textual descriptions, raising an important question: can LLMs reliably implement agent-based models from standardized s...

New preprint out. Can LLMs can go beyond "vibe coding" and generate scalable, maintainable agent-based models from formal specs?

🐑🐺Same ODD description, same predator-prey model. Results were… very interesting. arXiv 👇

arxiv.org/abs/2602.10140

Frontiers | Territorial and Spatial-Based Simulation Territorial and spatial-based modeling and simulation can provide valuable insights into various real-world systems, for example, by highlighting spatial pat...

Our research topic "Territorial and Spatial-Based Simulation" at @frontiersin.bsky.social is accepting submissions until May 31, 2026. Interest in submitting? Get in touch!

#territorialmodeling #spatialsimulation #agentbasedmodeling #agentbasedsimulation

www.frontiersin.org/research-top...

Structural Tree Extraction from 3D Surfaces This paper introduces a method to extract a hierarchical tree representation from 3D unorganized polygonal data. The proposed approach first extracts a graph representation of the surface, which serve...

The latest preprint by Diogo Andrade proposes a new method to extract tree-like structures from #3Dsurfaces for smarter #proceduralgeneration and #spatialanalysis.

#ULusofona #PCG #computationalgeometry #games #computergames #videogames #steinertree #🎮

arxiv.org/abs/2510.15886

ULusófona Researcher Defends Doctoral Thesis with Pioneering Contribution to AI | Universidade Lusófona Unanimously approved, the thesis is the eighth in the PhD in Informatics at Lusófona University – Lisbon University Centre.

🎓Congrats to our (now former) PhD student Alina Petukhova on a brilliant thesis defense!

Many thanks to our colleagues from @upv.es, @istecnico.bsky.social, and Lusófona University, for their roles on the jury and their rigorous and constructive feedback.

www.ulusofona.pt/en/news/ulus...

GPT-4.1 sets a new benchmark: reliably generating Python code for tough experiments with little-known libraries, while revealing gaps in docs. 📄

#AI #Python #Research #GPT41 @ULusofona @UNINOVA @lasige #codegeneration #codequality #LLMs

www.mdpi.com/1999-5903/17...

A simple question about our robot’s cleanup turned into a deep dive… and the definition of a new discrete statistical distribution 🤓🗑️➡️📊

Not world-changing, but definitely a fun journey with great colleagues!

👌U. Lusófona CTS/UNINOVA, FCUL, LASIGE, FCT

www.sciencedirect.com/science/arti...

Generating 3D Terrain with 2D Cellular Automata This paper explores the use of 2D cellular automata (CA) to generate 3D terrains through a simple additive approach. Experimenting with multiple CA transition rules produced aesthetically interesting,...

Open access version: arxiv.org/abs/2406.00443

Generating 3D Terrain with 2D Cellular Automata This paper explores the use of 2D cellular automata (CA) to generate 3D terrains through a simple additive approach. Experimenting with multiple CA transition rules produced aesthetically interesting,...

⛰️ Realistic landscapes can be generated by stacking iterations of 2D #cellularautomata into heightmaps! New paper to be presented at #IEEECoG next week.

💪With António Rodrigues, Diogo de Andrade, and @phil-lopes.com

👌 #ULusofona, #HEILab, #CTS/#UNINOVA, #FCT

ieeexplore.ieee.org/document/111...

Liminal Procedural Horror This article introduces a method for exploring liminal horror environments using a modified Wave Function Collapse algorithm. This approach employs real-time generated regions to create infinite, non-...

👹 New paper by Diogo Andrade where we propose a modified #WFC algorithm to generate unpredictable liminal #horror environments that heighten player unease. Diogo will be presenting it next week at #IEEECoG 2025 at @istecnico.bsky.social in #Lisbon.

ieeexplore.ieee.org/document/111...

GPT-4.1 Sets the Standard in Automated Experiment Design Using Novel Python Libraries Large Language Models (LLMs) have advanced rapidly as tools for automating code generation in scientific research, yet their ability to interpret and use unfamiliar Python APIs for complex computation...

New preprint: Can #LLMs run computational experiments?

✅ #GPT41 nailed it
⚖️ #Grok3 & #Mistral did well
❌ Smaller local models couldn't keep up

Tks: @ULusofona, @UNINOVA, FCUL, @lasige, @fct_pt, ILIND

arxiv.org/abs/2508.00033

pyclugen Multidimensional cluster generation in Python

New version 1.1.4 of PyClugen, your multidimensional #Python synthetic #cluster generator, available on @pypi.org with a very small #bugfix.

#syntheticdata #clustering #clusteringalgorithms #multidimensionaldata #unsupervisedlearning

pypi.org/project/pycl...

New paper out! It's an overview of deep learning methods for 2D semantic segmentation with benchmarks, datasets, metrics, and insights into future directions.
Acks ➡️ Lusófona University, FCT, CTS/UNINOVA, LASIGE/FCUL
#semanticsegmentation #deeplearning #2dsegmentation
www.mdpi.com/2673-4117/6/...

Detection and Characterization of Plume-Dominated Wildfires Extreme wildfires are increasingly hazardous, particularly plume-dominated fires, which exhibit unpredictable behavior due to their self-sustaining convective columns. Despite their significance, thes...

📝 New paper by our PhD student Afonso Oliveira in #DoCEIS25: "Detection and Characterization of PlumeDominated Wildfires" 🔥

Thanks for the support: U. Lusófona, Fundação para a Ciência e a Tecnologia, CTS/UNINOVA, LASIGE/FCUL!

#wildfires #plumefires #fireplumes

link.springer.com/chapter/10.1...