Comunidad Cenia How To
Andrés Abeliuk

Andrés Abeliuk

Especialidad: Redes sociales, procesamiento de lenguaje natural, teoría de la complejidad, ciencias sociales computacionales, machine learning.
Email: aabeliuk@dcc.uchile.cl
Andrés Abeliuk es Ph.D. in Computer Science por The University of Melbourne, título que obtuvo en 2017. Ha trabajado como Co-investigador en el proyecto “Analyzing Digital Menu Data to Characterize Nutritional Quality of Food Environments in Latino Neighborhoods”, parte del Southern California Center for Chronic Health Disparities in Latino Children and Families Pilot Projects Program. También es Investigador colaborador en la INICIATIVA CIENTÍFICA MILENIO, específicamente en el Instituto Milenio de Investigación sobre los Fundamentos de los Datos.

PUBLICACIONES

This article presents an exploration of the use of pre-trained Large Language Models (LLMs) for crisis classification to address labeled data dependency issues. We present a methodology that enhances open LLMs through fine-tuning, creating zero-shot and few-shot classifiers that approach traditional supervised models in classifying crisis-related messages. A comparative study evaluates crisis classification tasks using general domain pre-trained LLMs, crisis-specific LLMs, and traditional supervised learning methods, establishing a benchmark in the field. Our task-specific fine-tuned Llama model achieved a 69% macro F1 score in classifying humanitarian information—a remarkable 26% improvement compared to the Llama baseline, even with limited training data. Moreover, it outperformed ChatGPT4 by 3% in macro F1. This improvement increased to 71% macro F1 when fine-tuning Llama with multitask data. For the binary classification of messages as related vs. not related to crises, we observed that pre-trained LLMs, such as Llama 2 and ChatGPT4, performed well without fine-tuning, achieving an 87% macro F1 score with ChatGPT4. This research expands our knowledge of how to exploit the potential of LLMs for crisis classification, representing a great opportunity for crisis scenarios that lack labeled data. The findings emphasize the potential of LLMs in crisis informatics to address cold start challenges, especially critical in the initial phases of a disaster, while also showcasing their capacity to attain high accuracy even with limited training data.
Investigación Cenia 2025
Ir a la publicación

This article presents an exploration of the use of pre-trained Large Language Models (LLMs) for crisis classification to address labeled data dependency issues. We present a methodology that enhances open LLMs through fine-tuning, creating zero-shot and few-shot classifiers that approach traditional supervised models in classifying crisis-related messages. A comparative study evaluates crisis classification tasks using general domain pre-trained LLMs, crisis-specific LLMs, and traditional supervised learning methods, establishing a benchmark in the field. Our task-specific fine-tuned Llama model achieved a 69% macro F1 score in classifying humanitarian information–a remarkable 26% improvement compared to the Llama baseline, even with limited training data. Moreover, it outperformed ChatGPT4 by 3% in macro F1. This improvement increased to 71% macro F1 when fine-tuning Llama with multitask data. For the binary classification of messages as related vs. not related to crises, we observed that pre-trained LLMs, such as Llama 2 and ChatGPT4, performed well without fine-tuning, achieving an 87% macro F1 score with ChatGPT4. This research expands our knowledge of how to exploit the potential of LLMs for crisis classification, representing a great opportunity for crisis scenarios that lack labeled data. The findings emphasize the potential of LLMs in crisis informatics to address cold start challenges, especially critical in the initial phases of a disaster, while also showcasing their capacity to attain high accuracy even with limited training data.
Investigación Cenia 2026
Ir a la publicación

Publisher: ACM Transactions on Economics and Computation Link>

ABSTRACT

Algorithmic matching is a pervasive mechanism in our social lives and is becoming a major medium through which people find romantic partners and potential spouses. However, romantic matching markets pose a principal-agent problem with the potential for moral hazard. The agent’s (or system’s) interest is to maximize the use of the matching website, while the principal’s (or user’s) interest is to find the best possible match. This creates a conflict of interest: the optimal matching of users may not be aligned with the platform’s goal of maximizing engagement, as it could lead to long-term relationships and fewer users using the site over time. Here, we borrow the notion of price-of-anarchy from game theory to quantify the decrease in social efficiency of online algorithmic matching sites where engagement is in tension with user utility. We derive theoretical bounds on the price-of-anarchy and show that it can be bounded by a constant that does not depend on the number of users in the system. This suggests that as online matching sites grow, their potential benefits scale up without sacrificing social efficiency. Further, we conducted experiments with human subjects in a matching market and compared the social welfare achieved by an optimal matching service against a self-interested matching algorithm. We show that introducing competition among matching sites aligns the self-interested behavior of platform designers with their users and increases social efficiency.

RL5 2023
Ir a la publicación

Large Language Models (LLMs) can generate realistic text resembling human-produced content. However, the ability of these models to simulate conversations on social media is still less explored. To investigate the potential and limitations of simulated text in this domain, we introduce network-simulator, a system to simulate conversations on social media. First, we simulate the macro structure of a conversation using Agent-Based Modeling (ABM). The generated structure defines who interacts with whom, the type of interaction, and the agent’s stance on the topic of the conversation. Subsequently, using the simulated interaction structure, our system generates prompts conditioned on the simulation variables, producing texts that are conditioned on the parameters of the predefined structure, guiding a micro simulation process. We compare human conversations with those simulated by our system. Based on stylistic and model-based metrics, we found that our system can simulate conversations on social media in various dimensions. However, we detected differences in metrics related to the predictability of text production. Furthermore, we examine the effect of true and false framings within simulated conversations, revealing that simulated discussions surrounding false information exhibit a more negative collective sentiment bias than those based on true content.
RL3 2025
Ir a la publicación
agencia nacional de investigación y desarrollo
info@cenia.cl
Edificio de Innovación UC, Piso 2
Vicuña Mackenna 4860
Macul, Chile