Kleyton da Costa
I am a Ph.D. candidate in Computer Science at University College London and Senior Research Scientist at Holistic AI.
At UCL I am advised by Prof. Philip Treleaven and Prof. Dimitrios Kanoulas, where I work on predictive safety in embodied AI. My approach uses action-conditioned world models and hierarchical planning to anticipate and prevent unsafe behaviour in physical and decision-making agents.
My research interests include embodied AI, world models, graph learning, optimization, AI safety, and AI for science.
(Last update: September 2026)
News
- Oct 2026Paper ApprovedOur paper Automata from Agent Traces: Failure and Next-Step Prediction was accepted at NeurIPS 2026.
- Oct 2026Paper ApprovedOur paper GraphNetz: Statistical Benchmarking of Graph Neural Networks with Paired Tests and Rank Aggregation was accepted at the Learning on Graphs (LoG) Conference.
- Oct 2026Paper ApprovedOur paper Perspectives on Tsallis Statistics for Artificial Intelligence was accepted at Physica A.
- Oct 2026New RoleBecame a Senior Research Scientist at Holistic AI, working on local models and safe embodied AI.
- Jun 2026EducationStarted a Ph.D. in Computer Science at University College London.
Research
Representative papers are highlighted. Full list on Google Scholar.
Recent
Perspectives on Tsallis statistics for artificial intelligence
Kleyton da Costa, Bernardo Modenesi.
[Physica A: Statistical Mechanics and its Applications 2026]
GraphNetz: Statistical Benchmarking of Graph Neural Networks with Paired Tests and Rank Aggregation
Kleyton da Costa, Bernardo Modenesi.
[LoG Conference 2026] [PDF]
Generalized Graph Variational Autoencoders: Bounded Divergences Control Posterior Collapse
Kleyton da Costa, Bernardo Modenesi, Ivan F. M. Menezes, Helio Lopes.
[ArXiv 2026]
Automata from Agent Traces: Failure and Next-Step Prediction
Seonglae Cho, Franklin Cardenoso Fernandez, Umar Mohammed, Zekun Wu, Kleyton da Costa, Ilham Wicaksono, Adriano Koshiyama.
[Neurips 2026 (poster)] [PDF]
When Should Graph Attention Be Sparse? Learning a Per-Edge Tsallis Index
Kleyton da Costa, Bernardo Modenesi.
[ArXiv 2026] [PDF]
Are Single-Token Sparse Autoencoder Features Causally Necessary? Layer-Depth and SAE-Family Effects
Seonglae Cho, Zekun Wu, Kleyton da Costa, Rishi Kalra, Ilham Wicaksono, Adriano Koshiyama.
[ArXiv 2026]
A Survey on Predictive Safety in Embodied AI
Kleyton da Costa, Adriano Koshiyama, Dimitrios Kanoulas, Philip Treleaven.
[SSRN 2026] [PDF]
Divergence-Guided Particle Swarm Optimization
Kleyton da Costa, Bernardo Modenesi, Ivan F. M. Menezes, Hélio Lopes.
[ArXiv 2026]
The Confidence Manifold: Geometric Structure of Correctness Representations in Language Models
Seonglae Cho, Zekun Wu, Kleyton da Costa, Adriano Koshiyama.
[ArXiv 2026]
Time series forecasting via integrating a filtering method: an application to electricity consumption
Felipe Leite Coelho da Silva, Josiane da Silva Cordeiro, Kleyton da Costa, Nemias Saboya, Paulo Canas Rodrigues, Javier Linkolk Lopez-Gonzales.
[Computational Statistics 2025]
Learning on Graphs via Generalized Divergence Measures
Kleyton da Costa.
[M.Sc. Thesis, PUC-Rio 2025]
LibVulnWatch: A Deep Assessment Agent System and Leaderboard for Uncovering Hidden Vulnerabilities in Open-Source AI Libraries
Zekun Wu, Seonglae Cho, Umar Mohammed, Cristian Munoz, Kleyton da Costa, Xin Guan, Theo King, Ze Wang, Emre Kazim, Adriano Koshiyama.
[ACL 2025]
Evaluating explainability for machine learning predictions using model-agnostic metrics
Cristian Munoz, Kleyton da Costa, Bernardo Modenesi, Adriano Koshiyama.
[ArXiv 2024] [PDF]
Deep learning model fragility and implications for financial stability and regulation
Rishabh Kumar, Adriano Koshiyama, Kleyton da Costa, Nigel Kingsman, Marvin Tewarrie, Emre Kazim, Arunita Roy, Philip Treleaven, Zac Lovell.
[Bank of England 2023]
Solid Waste Management in Peru’s Cities: A Clustering Approach for an Andean District
Katherine Quispe, Mayra Martínez, Kleyton da Costa, Hilario Romero Giron, José Francisco Via y Rada Vittes, Lizet Doriela Mantari Mincami, Mohamed Mehdi Hadi Mohamed, Alex Rubén Huamán De La Cruz, Javier Linkolk López-Gonzales.
[Applied Sci. 2023]
Anomaly Detection in Global Financial Markets with Graph Neural Networks and Nonextensive Entropy
Kleyton da Costa.
[ArXiv 2023]
Statistical and Artificial Neural Networks Models for Electricity Consumption Forecasting in the Brazilian Industrial Sector
Felipe Leite Coelho da Silva, Kleyton da Costa, Paulo Canas Rodrigues, Rodrigo Salas, Javier Linkolk López-Gonzales.
[Energies 2022]
A machine learning approach to analyse ozone concentration in metropolitan area of Lima, Peru
Natalí Carbo-Bustinza, Marisol Belmonte, Vasti Jimenez, Paula Montalban, Magiory Rivera, Fredi Gutiérrez Martínez, Mohamed Mehdi Hadi Mohamed, Alex Rubén Huamán De La Cruz, Kleyton da Costa, Javier Linkolk López-Gonzales.
[Nature 2022]
