September 22, 2025
“To try something new” is how undergraduate student César Guerra-Solano (SCI ‘26) defines the word ‘innovate’. For him, innovation isn’t necessarily confined to a classroom or lab, rather a mindset he implements in all aspects of his lifestyle, especially research. A crucial component to this mindset is also connection and forging collaborations and welcoming perspectives from various backgrounds.
With this spirit of curiosity and problem-solving, Guerra-Solano sought out Professor Lorraine Li, a faculty member with SCI’s Department of Computer Science (CS), after learning about her research on the CS website. Together they began exploring abstract reasoning, the kind of pattern-based thinking that is a hallmark of human intelligence and a vital skill for AI systems. Inspired by the New York Times’ Connections game, Guerra-Solano proposed adapting the puzzle into a multilingual, culture-specific word grouping game, giving large language models (LLM’s) the task of identifying patterns and generating explanations for their groupings.
While the specifics of this game are straightforward in design, they are technically demanding in execution. To bring this idea to life, Guerra-Solano relied on the technical fundamentals he gained through SCI’s CS program and past internships.
“If it weren’t for hands-on class experience in some of the higher-level electives, like checking documentation, organizing my code, or the data science fundamentals from CS1656, I don’t know where I’d be!” Guerra-Solano said.
Prototypes And Challenges
Throughout the project, Python was essential, particularly the use of packages like Pandas for data analysis. His scripts for model evaluation and dataset analysis went through countless iterations, as he continually refactored them to improve efficiency, readability, and modularity. This process not only made it easier to integrate new functionality and analyses but also ensured that anyone with Python experience could follow his work. This proved to be invaluable when writing about the project in research papers or collaborating with new team members.
“I didn’t come up with this project,” Dr. Li emphasized. “César not only proposed the task as an evaluation method but also led the majority of the coding, analyzing datasets, categorizing game types by difficulty, and testing prompts across closed-source models”.
But transforming a casual puzzle into a rigorous research tool was no small feat. Guerra-Solano describes the biggest challenge as getting started with each “large” component of the project: designing experiments, refining the game, coding the scripts, and writing up the results.
“There was definitely a learning curve,” he said, “but thanks to the mentorship of Dr. Li and PhD student Zhuochun Li, I gained skills I never had before. Trial and error combined with their feedback let me grow more independent as a researcher.”
For Guerra-Solano, curiosity about large language models is rooted in a lifelong fascination with language. Growing up bilingual, he saw language as both a bridge to culture and a powerful means of connection and expression. That interest deepened in high school, when Guerra-Solano worked on a summer natural language processing (NLP) research project exploring racial bias in the Google Ngram program. His perspective has also been shaped by seeing the potential of tools like ChatGPT, neuroscience courses, and work experiences, which taught him the value of understanding underlying mechanisms, whether in the brain or in AI systems.
“As an undergraduate researcher, I’ve been blessed with the opportunity to interact with unfamiliar and exciting problems… I get to tackle and learn something new,” Guerra-Solano noted.
Leadership and Teamwork
Guerra-Solano’s curiosity and independence are matched by his ability to lead and collaborate effectively. Dr. Li emphasizes that Guerra-Solano’s approach has been integral to the project’s success: “César’s ability to communicate and onboard new members has been crucial. He consistently identifies novel ideas, designs creative experiments, and presents his results with clarity. His contributions have been the cornerstone of this work.”
Now, Guerra-Solano’s curiosity is extending beyond the lab. Together with Julie Lawler (SCI '25), he is working to start a tech club coined Technology 4 Good with Dr. Li serving as an advisor. The idea grew out of Guerra-Solano’s involvement of the club All Tech is Human (ATIH), where he and the founder recognized the urgent need for open discussions on AI ethics, particularly as LLMs and other AI technologies became more widespread.
“I have the dream that in a perfect state, [AI language systems] can transform the world, from being aids for students, to being AI assistants that can improve accessibility to things no matter someone’s background. The challenge is getting there.”
For SCI students, opportunities like César’s provide more than just a space to sharpen problem-solving and creativity but also highlight the importance of connection across faculty, peers, and the broader community. By engaging with research early, students can discover new passions, build meaningful relationships, and experience firsthand how it can inspire, challenge, and change the world.