Check it regularly, as the schedule for keynote speakers may change slightly.
CST - Central Time Zone (Time in Lima City, Lima, Peru) | * Presenting author
| Time | Title | Author(s) |
|---|---|---|
| 09h15 - 09h30 | Reception | |
| 09h30 - 09h45 | Welcome to SIMBig 2026 — Authorities of the UCSP and the Co-Chairs of SIMBig 2026 | |
| Session 1: Applied NLP and Language Model Applications | ||
| 09h45 - 10h00 | Search and Rescue: Benchmarking Web-Based Retrieval in Modern LLMs | Kale, Sahil* |
| 10h00 - 10h15 | Where LLM Knowledge Belongs: A Controlled Study of Semantic Injection Surfaces in SASRec-Style Sequential Recommendation | Wizenfeld, Jeremy; Bouneffouf, Djallel* |
| 10h15 - 10h30 | Gender Bias Mitigation with Activation Steering and Fine-Tuning LLMs | Zhou, Hanqing*; Inkpen, Diana; Kantarci, Burak |
| 10h30 - 11h30 |
Invited Speaker: Dan JURAFSKY Stanford University, USA |
 |
| The Social Failures of Language Models as Conversational Partners: People converse with language models for information, advice, and emotional support. In this talk I'll summarize studies in our lab showing that models fail in systematic ways as social interlocutors. We find that language models are socially sycophantic, linguistically overconfident, overly anthropomorphic, and epistemically self-centered. We then show that these flaws have real consequences for users: people interacting with models suffer consequences including overreliance, distorted judgment, and reduced personal responsibility. I'll discuss datasets and metrics, explore mitigations, and call for design, evaluation, and accountability mechanisms to protect user well-being. | ||
| 11h30 - 11h45 | Coffee Break 15 min | |
| 11h45 - 12h00 | Personality-Aware Transformer Features for Suicide-Note Analysis | Lazo Vasquez, Ricardo*; Ochoa Luna, José |
| 12h00 - 12h15 | Beyond Student Labels: Evaluating Prompt Sensitivity in AI-Generated Practice Problems for Multilingual Learners | Rudraraju, Satyanarayana* |
| 12h15 - 12h30 | Few-Shot Requirements Classification: An Empirical Evaluation of MPNet and BETO on English and Spanish Datasets | Mogollon, Sergio* |
| 12h30 - 12h45 | Semantic Review Mining for Regulatory Risk Detection | Chianumba, Ernest; Santoni, Jesse; Samuel, Raina* |
| 12h45 - 13h00 | Transformer-Based Models for Health Information Disorder Analysis: Misinformation Detection and Propagation Prediction | Alvarado Chavez, Margiory*; Manuico-Quequejana, Flavia; Pontalti Monari, Ana Carolina; Soriano-Vargas, Aurea |
| 13h00 - 14h00 | Lunch 1 hour | |
| 14h00 - 14h15 | Transcription of Historical Documents using Seq2Seq Models | Laguna, Rodrigo*; Chavat, Felipe; Guillermo Moncecchi, Guillermo; Garat, Diego |
| 14h15 - 14h30 | Detecting and Mitigating Deception in Multilingual SLMs during Spanish-Language Tasks via Linear Probes | Sirena, Sergio* |
| Session 2: Image Processing and Computer Vision Applications | ||
| 14h30 - 14h45 | Instance Segmentation of Defects in Hand-Spun Alpaca Fiber Yarn Using YOLO Models | Montoya Cubas, Carlos Fernando*; Hancco Champi, Fran Anthony; Mamani Meza, Johana Maria; Achahui Cruz, Milton Amed; Díaz Jurado, José Mauricio; Berrocal Cueto, Jhordan |
| 14h45 - 15h00 | Visual Analytics for the Interpretable Multimodal Analysis of Plant Diseases using Vision-Language Transformers | Diaz, Berly* |
| 15h00 - 16h00 |
Invited Speaker: Gabriela CSURKA NAVER LABS, France |
 |
| Towards 3D Foundational Models with Spatial intelligence: In this talk I would like to take you on a journey through our development from CroCo to DUSt3R, which has brought a paradigm shift in 3D vision unifying the streamlined processing of 3D visual data. By exploring derivative and follow-up work from both our team and the broader community, I will highlight how this framework has opened new avenues for solving complex 3D challenges like map-free visual localization, robotic navigation, and beyond. Next, I will discuss it's extension towards different class of multi-view 3D reconstructions models using feed-forward transformer such as MUSt3R, Spann3R, CUT3R and VGGT enabling direct reconstruction of large, multi-view datasets. I’ll continue with PanSt3R, a system built on MUST3R, that incorporates semantic understanding, allowing for joint 3D reconstruction and panoptic segmentation directly from unposed 2D image collections. By combining geometric reasoning with semantic perception, PanSt3R represents a significant step toward end-to-end 3D scene understanding from raw imagery. Finally, I will conclude my talk by discussing how to combine such 3D Foundational Models with LLMs to enhance their scene understanding and spatial intelligence. | ||
| 16h00 - 16h15 | Coffee Break 15 min | |
| 16h15 - 16h30 | From Monocular Video to 3D Pose and Kinematic Descriptors: A Padel-Oriented Pipeline with Limited 3D Data | Corral, Alejo*; Dominguez, Martin; Molina, Matias |
| 16h30 - 16h45 | 3D Spatial Byte-Pair Encoding: Graph-Contraction Tokenization for Volumetric CT | Thota, Surya*; Saharia, Navanath |
| 16h45 - 17h00 | Image–Physics Complementarity for M- and X-Class Solar Flare Prediction: A Multimodal Study with Swin3D and BiLSTM | Muñoz, Giomar*; Cardenas, Rolando |
| 17h00 - 17h15 | Spatiotemporal Prediction of Land Use and Land Cover Change in Cusco Using Random Forest, Cellular Automata and Markov Chains | Choque-Navarro, Marcelo Fabian*; Medrano-Valencia, Ivan Cesar; Vera-Olivera, Harley; Alvarez-Mamani, Edwin |
| Session 3: Applied Machine Learning for Society | ||
| 17h15 - 17h30 | Riemannian Factors for Forecasting Costa Rican Inflation: A Comparative Analysis with PCA and KPCA in FAVAR Models | Benavides Castillo, Dylan*; Asch, Christian; Cuendis, Jostin; Rodríguez, Oldemar |
| 17h30 - 17h45 | Evaluating Climate Representations for Agricultural Productivity Modeling: Evidence from Peru | Urrea Rodriguez, Clara Patricia*; Villavicencio Lopez, Luis Felipe; Nuñez Del Prado Cortez, Miguel |
| 17h45 - 18h00 | Do TFMs Reduce the Need for Statistical Preprocessing in Small-Sample Excess-Zero Environmental Microbiome Data? | Orellana, Omar; Valenzuela, Luis; Marti, Luis; Chauvin, Gautier; Enaud, Raphael; Delhaes, Laurence; Avalos-Fernandez, Marta* |
| 18h00 - 18h15 | Latency-Aware Streaming Predictive Architecture for Silica Estimation in Iron Ore Flotation Using Machine Learning, Apache Kafka, and Digital-Twin-Style Monitoring | Puruguay, Sebastian*; Meza, Alexis; Vilcatoma, Joel; Calderon, Marks |
| 18h15 - 18h30 | Segment-Emergent Association Rules: Separating Temporal Signals from Partitioning Artifacts | Santisteban, Julio*; Arevalo Alvis, Jorge |
| 18h30 - 18h45 | Explainable Machine Learning and Spatial Equity Analysis of University Admission Outcomes in the Puno Region, Peru | Cutipa-Santi, Michael Newton*; Quelca-Velasquez, Daisy Crishyely; Cutipa-Santi, Yeni; Mendoza-Mollocondo, Charles Ignacio; Tito Lipa, José Pánfilo |
| 18h45 - 19h00 | 3-Phase Multi-RAG Architecture: Optimization of Routing and Information Retrieval in University Admissions Processes | Arroyo-Paz, Antonio*; Chambi-Velasquez, Rommel; Hilacondo-Begazo, Andre |
| Time | Title | Author(s) |
|---|---|---|
| 09h15 - 09h30 | Reception | |
| Session 1: Machine Learning Fundamentals | ||
| 09h30 - 09h45 | Physics-Informed Graph Neural Networks for Missing Data Imputation in Unbalanced Three-Phase Distribution Grids | Peña Torres, Josemanuel Alejandro; Quispe Zavala, Rosmery Violeta; Gutierrez-Pachas, Daniel Alexis* |
| 09h45 - 10h00 | Adaptive Semantic Anchors: Persistent Geometric Regularization for Cold-Start Matrix Factorization | Ohayon, Ariel; Bouneffouf, Djallel* |
| 10h00 - 11h00 |
Invited Speaker: Tom M. MITCHELL Carnegie Mellon University, USA |
 |
| The History of Machine Learning: Machine learning is the key technology underlying today's amazing artificial intelligence systems. How did we get to today's technology which now supports a trillion dollar AI industry? What were the key scientific breakthroughs? What were the surprises and dead-ends along the way, as seen by the researchers who created them? Who were the personalities involved, and what were they thinking at the time? What should we learn from all this? This talk will explore the history of machine learning based on personal experience of the speaker, augmented by video interviews with a dozen other pioneering researchers in the field taken from the new podcast "Machine Learning: How Did We Get Here?" | ||
| 11h00 - 11h15 | Coffee Break 15 min | |
| 11h15 - 11h30 | I-SEL: Boosting Learning Algorithms with Interval-Valued Statistically Enhanced Learning | Benavides Castillo, Dylan* |
| 11h30 - 11h45 | Offline Metrics for Explanations over Continuous Latent Factors | Espinoza, Edgar*; Sarmiento, Edgar |
| 11h45 - 12h00 | A quantum benchmark case for Multi-swarm and quantum PSO | Ferrante, Agustin*; Dominguez, Martin; Osenda, Omar |
| Session 2: Health Applications and Biomedicine | ||
| 12h00 - 13h00 |
Invited Speaker: Hoifung POON Microsoft Research, USA |
 |
| Toward Virtual Patient: AI for Accelerating Biomedical Discovery: Human biology stems from four billion years of evolution, which creates a proliferation of complexity and heterogeneity. The combinatorial explosion of biological hypotheses easily surpasses the number of atoms in the universe. Meanwhile, biomedical discovery advances at the snail pace of clinical trials, each taking years and costing over $100 million. As we enter the era of precision health and try to tailor treatments for each individual, continuing on today's discovery processes is clearly unsustainable. Our overarching research agenda lies in advancing multimodal generative AI to create a biomedical world model by bridging virtual patient and virtual cell, thus turning real-world patient journeys into a discovery engine through a generalized lab-in-the-loop flywheel. This enables us to synthesize population-scale real-world evidence from hundreds of millions of patients and accelerate biomedical discovery through AI-powered virtual clinical trials. This in turn fuels diverse applications from imminent productivity gains via automating digitized knowledge work, to moonshot creativity gains via creating biomedical wind tunnels for in silico simulation. By leveraging AI to close the loop across clinical care and frontier research, we have a clear line of sight toward scalable drug discovery. | ||
| 13h00 - 14h00 | Lunch 1 hour | |
| 14h00 - 14h15 | Machine Learning and Multi-Omic Enrichment Analysis for Breast Cancer Classification from mRNA and miRNA Microarray Profiles | Ramos Alvarez, Ismael Gerson*; Palma Ttito, Luis Beltran; Alzamora Paredes, Robert Wilbert; Rozas Huacho, Javier Arturo; Montoya Cubas, Carlos Fernando |
| 14h15 - 14h30 | Multimodal Brain Age Estimation with a beta-Variational Autoencoder and Gradient Boosting on a Multicentric Latin American Cohort | Barsotti, Damián*; Bazán, Nicolás M.; Della Bella, Gabriel |
| 14h30 - 14h45 | MamaDetect AI: A Multi-Task Deep Learning Pipeline for Classification, Detection, and Segmentation of Breast Masses in DICOM Mammography | Bravo, Yessenia*; Meza Morveli, Frayder; Espinoza Lapa, Shirel; Sánchez Toro, Yanely; Fernández Mejía, José; Fabian Arteaga, Junior |
| 14h45 - 15h00 | Improving Lightweight Models for Lung Cancer Detection via Two-Stage Knowledge Distillation | Navarro Guillen, Emerson*; Yauri Vidalon, Jose Elias; Vargas Sotaya, Yakelin; Zapata Casaverde, Richard; Fernández Jerí, Lissette Elvira; Luján Pérez, Diana |
| 15h00 - 16h00 |
Invited Speaker: Hugo AERTS Harvard University, USA |
 |
| Beyond Prediction: Using AI to Discover Hidden Biology in Medical Images: Artificial intelligence in medicine is often framed as a tool for automating clinical tasks or predicting patient outcomes. But its greater potential may be to reveal biological information that has remained hidden within routinely collected medical data. In this talk, I will discuss how advances in foundation models and multimodal AI are enabling medical images to be interrogated at scale, turning them into quantitative measurements of human biology. I will highlight recent work showing how AI analysis of routine CT imaging can measure thymic health in adults and uncover its relationships with cancer risk, survival, immune function, and response to cancer therapy. These studies illustrate a broader opportunity: moving beyond models that simply predict outcomes toward AI systems that generate biological insight, identify new biomarkers, and ultimately help guide prevention and treatment. I will discuss the opportunities and challenges of translating these discoveries into clinically useful tools and how this emerging paradigm could reshape the role of AI in biomedical research. | ||
| 16h00 - 16h15 | Coffee Break 15 min | |
| 16h15 - 16h30 | Machine Learning-Based Classification of Dengue Severity in Peru Using Public Health Data | Flores-Carbajal, Luz M.; Gonzales-Castillo, Julio C.; Mamani-Coaquira, Yonatan* |
| 16h30 - 16h45 | How Should Emergency Department Resource Utilization Be Clustered? A Comparative Study of Distance Metrics and Clustering Strategies | Elorga-Castagnet, Nadia*; Guerra-Adames, Ariel; Russon, Dylan; Doremus, Océane; Gil-Jardiné, Cédric; Avalos-Ferandez, Marta |
| Session 3: Ontologies, Semantic Search and Retrieval | ||
| 16h45 - 17h00 | Contrastive Learning for Semantic Retrieval of Peruvian News Articles: A Comparative Evaluation Using a Journalistic Corpus | Arroyo-Paz, Antonio*; Ezcurra-Paima, Maria; Marroquin-Pinto, Paola |
| 17h00 - 17h15 | HySOR: Hybrid Semantic-Ontological Retrieval for Aviation Safety Incident Analysis over Fused ASRS-NTSB Data | Paulraj, Vasantha Selvi*; Paulraj, Vasantha selvi; Madhavan, Manu |
| 17h15 - 17h30 | Co-occurrence Graph and Relevance-Based Filtering for Precision Improvement in CSR-RAG Table Retrieval | Apaza Iruri, Mauricio*; Villanueva Llerena, Julissa |
| 17h30 - 17h45 | Retrieval-Augmented Fine-Tuning of Small Language Models for Domain-Specific Technical Reasoning in Peru's Sanitation Sector | Moreno Alvarado, Herman*; Ochoa-Luna, José |
| Session 4: Machine Learning Fundamentals II | ||
| 17h45 - 18h00 | The Carbon Footprint of Optimizers in Deep Neural Networks: A Green AI Benchmark | Camino, Adrian Emilio; Navarrete, Dax*; Flores-Reyes, Alejandro; Cabrera, Danny; Suntaxi, Gabriela; Recalde, Lorena; Martinez-Mosquera, Diana |
| 18h00 - 18h15 | A Model-Agnostic Temporal Explainability Framework for Financial Time Series Prediction | Zavalaga, Rushell* |
| 18h15 - 18h30 | Exploring Around the Collapse of Learned Heuristics with Restricted Novelty | Maldonado Quispe, Percy*; Gaona Briceño, Leonardo Gustavo |
| 18h30 - 18h45 | Privacy-Preserving Distributed Neural Networks | Cáceres Romero, Sergio; Candido, Renato; Silva, Magno* |
| Time | Title | Author(s) |
|---|---|---|
| Session 1: Shared Task | ||
| 09h15 - 10h15 | Shared Task on ASR and Language Resources for Indigenous Languages of South America | Huaman, Elwin*; Córdoba, Johana* |
| Session 2: Data-Driven Systems and Intelligent Applications | ||
| 10h15 - 10h30 | Model Debt: The Hidden Cost of Multi-Model Vibe Coding in Production Software for First Responders — An Extended Case Study | Paul, Sonipriya*; Thomas, Dominic; Rijal, Aarya; Safiyari, Hamidreza |
| 10h30 - 10h45 | Towards Reactive and Intelligent Smart Laboratories: A Multi-Agent Context-Aware Software Architecture | Paucar Curasma, Herminio*; Cezar Estrella, Julio |
| 10h45 - 11h00 | Scaling Character Adherence: Efficiently Adapting Small Language Models for Interactive NPCs | Butron Cuayla, Daryl*; Lazo Vasquez, Ricardo Manuel; Ochoa Luna, José Eduardo |
| 11h00 - 11h15 | Coffee Break 15 min | |
| 11h15 - 11h30 | Semantic Search of Scientific Abstracts through HNSW Indexing in Vector Databases | Arroyo-Paz, Antonio*; Callocondo-Garay, Santiago; Mejia-Ramos, Piero |
| 11h30 - 11h45 | When Semantic Alignment Helps Sequential Recommendation | Wexler, David; Bouneffouf, Djallel* |
| 11h45 - 12h45 |
Invited Speaker: Omar ALONSO Amazon Science, USA |
 |
| Domain-specific knowledge graphs construction: challenges and opportunities: Knowledge graphs (KGs) are foundational infrastructure for search, question answering, and entity-centric applications. Recently, KGs are being used in many LLM-based applications. While there are examples of knowledge graphs in industry and academia, there is little information on the practical aspects of design and construction. The KG lifecycle, from construction over heterogeneous sources to evaluation at scale, raises data management problems that shape the reliability and utility of every system built on top of them. In this talk, we describe some of the challenges and opportunities when designing and implementing a domain-specific knowledge graph from the ground up. | ||
| 12h45 - 13h00 | Guidelines for the Design of an Artificial Intelligence-Based Mobile Application for Dietary Tracking and Gastrointestinal Pattern Recognition in University Students | Díaz Suárez, Jorge Eduardo*; Atencio, John; Frisancho, Sebastian; Iñiguez-Jarrin, Carlos |
| 13h00 - 14h00 | Lunch 1 hour | |
| 14h00 - 14h15 | Face Verification and Presentation Attack Detection for University Access Control | Vargas Saldaña, Daniella* |
| 14h15 - 14h30 | Data-Driven Geostatistical Optimization of LoRaWAN Sensor Placement for Urban Heat Island Monitoring in Piura, Peru | Ocaña Rufino, Jazmín; Samillán Vera, Ángel; La Rosa Lama, Gerson*; Soto Bohórquez, Juan |
| Session 3: Education, Society and Machine Learning Applications | ||
| 14h30 - 14h45 | Operationalizing Bias Mitigation in Scholarly Information Systems: Fairness-Utility Trade-offs across Prediction and Retrieval | Recalde, Lorena*; Sarango, César; Martinez-Mosquera, Diana; Suntaxi, Gabriela |
| 14h45 - 15h00 | A Web System with Artificial Intelligence and Gamification for Personalized Tutoring of Primary School Students in Peru | Garro, Cesar* |
| 15h00 - 15h15 | Sentiment Analysis in Collao Quechua under Limited Data: Lessons from an Automatically Annotated Corpus | Marín-Rodríguez, Sintia Vallet; Paccotacya-Yanque, Rosa* |
| 15h15 - 15h30 | Digital Reading and Writing Indicators for Early Dyslexia Risk Screening in Primary School Students: A Conceptual Mobile Model | Campos, Silvia*; Huamani, Joseph; Quispesivana, Claudio |
| 15h30 - 15h45 | What Survives an Uneven Record? Fine-Grained Crime Prediction under Reporting Bias: Evidence from Peru | Vilchez, Rody*; Velasquez, Christian; Alarcon, Carlos |
| 15h45 - 16h00 | Coffee Break 15 min | |
| 16h00 - 16h15 | An Empirical Benchmark of Machine Learning Models for Short-Term GNSS S4 Forecasting at Jicamarca, Peru | Ayque, Tania*; Nina, Rafael; Schreiber, Diego; Atencio, Carlos |
| 16h15 - 16h30 | Explainable Credit Risk Assessment with FT-Transformer and Multi-Agent LLMs | Maldonado Quispe, Percy*; Huanca Olazabal, Cristhian David |
| 16h30 - 16h45 | Unsupervised Learning for Profiling University Students with Parental Responsibilities | Cutipa-Santi, Yeni; Cutipa-Santi, Michael Newton*; Quelca-Velasquez, Daisy Crishyely; Apaza-Cutipa, Renzo; Ibañez-Quispe, Vladimiro |
| 16h45 - 17h00 | Multi-Level Visual Analytics of Online Shopper Behavior: K-Means++ and Digital Attention Metrics as an Eye-Tracking Proxy | Valdivia Mendoza, Juan Diego* |
| 17h00 - 17h15 | SIMBig 2026 Closing — Authorities of the UCSP and the Co-Chairs of SIMBig 2026 | |
