Published:
Names are one of the first things a system asks for, and one of the easiest ways it reveals what culture it was built for. This post looks at how seemingly harmless assumptions in software can quietly erase identity, using Chinese names as the primary case study.
Published:
How do you build real connections outside your home academic program or department, especially in a place like MIT? This post shares a simple playbook for doing exactly that, using my own path into the CS/AI community as a case study. The short version: find the right announcement surfaces, show up (even when you won’t understand everything), and keep the curiosity dial set to “loud.”
Published:
Hi there. If you wind up on this page, I assume you’re busy studying for your upcoming MBA Core exams at MIT Sloan.
Published:
Let’s setup MIT thesis LaTeX template locally!
Published:
This post is a quick guide to setting up LaTeX on your own computer, so you can write papers, resumes, or dazzling “class notes” using LaTeX locally—without relying on Internet connectivity. In other words, you won’t be tied to Overleaf: you can bring your laptop anywhere and keep working on your next big ideas.
Published:
Abstract: Vision-language models like CLIP struggle with multi-object scenes, often favoring prominent objects or those mentioned first in captions. Using real-world COCO images, we show that CLIP’s caption-matching accuracy drops from 91.23% to 87.45% when object order is reversed. To address this, we explore a post-hoc mitigation: a permutation ensemble that averages scores across all object orders, boosting robustness and recovering accuracy to 90.04%. Our findings reveal persistent order biases and offer a simple, effective strategy to improve CLIP’s reliability in complex scenes.