A well-designed poster is like a siren’s call for potential collaborators. Picture yourself strolling through a maze of posters during a conference tea break. Even with the best intentions to explore, you’re forced to make snap decisions. One poster catches your eye. You pause, then strike up a conversation with its creator. Mission accomplished—for both the poster and its owner. Posters actually pull double duty. First, they’re your sidekick during the pitch talk during the very poster session. Second—and often overlooked—they’re the academic wallpaper of conferences. Organizers love filling hallways with posters because they add that scholarly vibe. Attendees wander, glance, and absorb. This is where a stand-alone, easy-to-digest poster with clear contact info can work unexpected wonders. Who knows? You might land your next collaborator—or admirer—while they’re lost finding a bathroom. ...

The textbook Multi-armed Bandit Allocation Indices by John Gittins, Kevin Glazebrook, Richard Weber has a charming Chapter 6, titled “Restless bandits and Lagrangian relaxation”. But the authors skipped a step in the prove of how to go from the LP to Dual and to Whittle Index Policy. Prelims: An RMAB (Restless Multi-Armed Bandit) instance is defined through the tuple $$ > \lang S, A, r_{i}, \mathcal P_{i} \rang_{i \in [N]} > $$ where $S_i $ is the state space, $A_i = \lbrace 0, 1 \rbrace$ the action space, $r_{i}$ and $\mathcal P_{i}$ denote the reward and transition kernels arm $i$. ...

Happy weekend and winter solstice—if winter comes, will spring be far behind? Old CV is archived here.

Here’s a paper I read lately, Generative Social Choice Sara Fish, Paul Gölz, David Parkes, Ariel Procaccia, Gili Rusak, Itai Shapira, and Manual Wüthrich. link to one of the authors’ posted verison. Plus, link to ICML talk version. [Abstract] The mathematical study of voting, social choice theory, has traditionally only been applicable to choices among a few predetermined alternatives, but not to open-ended decisions such as collectively selecting a textual statement. We introduce generative social choice, a design methodology for open-ended democratic processes that combines the rigor of social choice theory with the capability of large language models to generate text and extrapolate preferences. Our framework divides the design of AI-augmented democratic processes into two components: first, proving that the process satisfies representation guarantees when given access to oracle queries; second, empirically validating that these queries can be approximately implemented using a large language model. We apply this framework to the problem of summarizing free-form opinions into a proportionally representative slate of opinion statements; specifically, we develop a democratic process with representation guarantees and use this process to portray the opinions of participants in a survey about abortion policy. In a trial with 100 representative US residents, we find that 84 out of 100 participants feel “excellently” or “exceptionally” represented by the slate of five statements we extracted. ...

The funny news (five years ago), original by Metro’s Basit Mahmood: Chinese businessman Tan Youhui was looking for a hitman to take out a competitor, Wei Mou, and was willing to pay 2 million yuan (£218,000) to get the job done. The hitman that Mr Youhui hired decided to offer the job to another hitman for half the original price. The second hitman then subcontracted to another hitman, who then subcontracted to a fourth, who gave the job to a fifth. ...

Two things that are notoriously slippery to pin down: fairness and democracy. Fairness—what is it, really? To me, fairness is just another way of thinking about efficiency. To elaborate, we often define efficiency as the sum of everyone’s value or surplus. But maybe that’s too simplistic. What if fairness is just a more complex version of that same idea? Instead of a straightforward sum, fairness might be a more intricate mapping of each participant’s value into the final outcome. In the end, achieving fairness feels a lot like maximizing some kind of welfare function—it’s just a matter of balancing competing interests and getting the best overall result. ...

The story starts with today’s law class. For the last topic of this semester, we dove into healthcare policies—And wow, the slides are a work of art: Yes, I’m flexing on my course materials. It is titled ‘A Glimpse into U.S. Society from Supreme Court Rulings’ Healthcare in the United States used to be (and might still be) a mess (but who’s doing good, anyway?) Going to the hospital is expensive, so most people don’t pay directly out of pocket and instead have it covered by private insurance or government programs. In some way, the market’s design in US has created a vicious cycle. ...

December 15th is the unofficial standard for PhD applications starting next fall, but of course, academia loves its quirks. Columbia Business School goes for December 16th, Carnegie Mellon wraps up by December 11th, and some elite schools close as early as December 1st. Some schools seem more relaxed, with deadlines stretching into January. Let’s be honest—I guess no one’s reading applications until after winter break anyway. Fast forward to the next big date: April 15th. This one often gets billed as a “decision deadline”—UCLA’s FAQ clears it up: ...

Supply chains are the unsung heroes of every industry—arguably as important as market rules, if not more so. And when it comes to China’s most popular drink, lemonade, the role of supply chains becomes even more crucial. Today’s story is of how MiXue, one of China’s most successful beverage brands, conquered the market with its mastery of supply chain innovation. MiXue dominates the low-price milk tea market in China, earning it the title of ‘Starbucks of low prices.’ ...

One of my coauthor shared this immensely inspiring advice with me: write the paper first by Jason Eisner (2010). In short: If you’re planning to submit a conference paper, I’d like to strongly suggest that you spend the next few days just writing the paper (even if you haven’t yet planned or finished the experiments). The motivations is explained from the perspective of getting computer science papers ready for submission to conferences: ...