Jasin Cekinmez
Hi, I’m Jasin, a senior at Princeton studying applied math and computer science. I build evaluations of multimodal AI systems, probing where they fail, why, and how their behavior diverges from humans.
Along the way, I’ve interned on the Query Optimization team at Teradata and as an ML researcher at QCRI. When I’m not stress-testing models, I’m probably constructing crosswords 🧩 for The New York Times.
Research
SportD: Can VLMs Physically Strategize?
Where Did the Ambiguity Go? Examining How Multimodal Models Interpret Polysemous Words
OccludeBench: Can VLMs See the Hint?
Query Timing Produces Opposite Positional Biases Between LLMs and Humans
ADAM: A Diverse Archive of Mankind for Evaluating and Enhancing LLMs in Biographical Reasoning
Guess the Unified Model: Domain and Linguistic Effects in Generated Images
Quantifying Empirical Compute-Supervision Tradeoffs in RLVR