My ballet class got rained out, so I ducked into a cozy café—and somehow ended up learning a bit more about Meituan and Ele.me’s business models over two cappuccinos—real coffee talk. In that café, when the barista handed the delivery guy one packed coffee. Intro: On these food delivery platforms, cafés are ranked for visibility. The higher the rank, the more orders roll in. And while it’s no surprise that paying to jump the line (read: sponsored listings) exists, what is surprising is how quietly the game is being played—not all money-boosted rankings are labeled. In fact, many aren’t. ...

Awesome paper I came across in WWW25’, using LLM to summarize a super note on top of all the community notes: Supernotes: Driving Consensus in Crowd-Sourced Fact-Checking (Courtesy to Soham’s website for the info) Soham De, Michiel A. Bakker, Jay Baxter, and Martin Saveski The ACM Web Conference. 2025 X’s Community Notes, a crowd-sourced fact-checking system, allows users to annotate potentially misleading posts. Notes rated as helpful by a diverse set of users are prominently displayed below the original post. While demonstrably effective at reducing misinformation’s impact when notes are displayed, there is an opportunity for notes to appear on many more posts: for 91% of posts where at least one note is proposed, no notes ultimately achieve sufficient support from diverse users to be shown on the platform. This motivates the development of Supernotes: AI-generated notes that synthesize information from several existing community notes and are written to foster consensus among a diverse set of users. Our framework uses an LLM to generate many diverse Supernote candidates from existing proposed notes. These candidates are then evaluated by a novel scoring model, trained on millions of historical Community Notes ratings, selecting candidates that are most likely to be rated helpful by a diverse set of users. To test our framework, we ran a human subjects experiment in which we asked participants to compare the Supernotes generated by our framework to the best existing community notes for 100 sample posts. We found that participants rated the Supernotes as significantly more helpful, and when asked to choose between the two, preferred the Supernotes 75.2% of the time. Participants also rated the Supernotes more favorably than the best existing notes on quality, clarity, coverage, context, and argumentativeness. Finally, in a follow-up experiment, we asked participants to compare the Supernotes against LLM-generated summaries and found that the participants rated the Supernotes significantly more helpful, demonstrating that both the LLM-based candidate generation and the consensus-driven scoring play crucial roles in creating notes that effectively build consensus among diverse users. ...

I watched the Australian Ballet’s 2025 season Manon at the Sydney Opera House. Mia Heathcote danced Manon tonight—she’s a soloist now, but her technique is amazing and I believe she will rise up to be principal soon. Manon’s story is unconventional for ballet where the heroine is usually, well, not a ‘slut’. The synopsis is centered around Manon, the young, beautiful but poor woman with nothing but pretty privilege, torn to choose between her love for the impoverished student Des Grieux and the allure of wealth of Monsieur G.M. Ultimately she is arrested and deported for being a prostitute, with Des Grieux following her to a penal colony and she died. ...

My advisor was busy during tea break so I was sent for the paper’s oral presentation. His loss ;) Behavior Modeling Space Reconstruction for E-Commerce Search Wang et al. WWW2025. https://doi.org/10.48550/arXiv.2501.18216 Delivering superior search services is crucial for enhancing customer experience and driving revenue growth. Conventionally, search systems model user behaviors by combining user preference and query item relevance statically, often through a fixed logical ‘and’ relationship. This paper reexamines existing approaches through a unified lens using both causal graphs and Venn diagrams, uncovering two prevalent yet significant issues: entangled preference and relevance effects, and a collapsed modeling space. To surmount these challenges, our research introduces a novel framework, DRP, which enhances search accuracy through two components to reconstruct the behavior modeling space. Specifically, we implement preference editing to proactively remove the relevance effect from preference predictions, yielding untainted user preferences. Additionally, we employ adaptive fusion, which dynamically adjusts fusion criteria to align with the varying patterns of relevance and preference, facilitating more nuanced and tailored behavior predictions within the reconstructed modeling space. Empirical validation on two public datasets and a proprietary search dataset underscores the superiority of our proposed methodology, demonstrating marked improvements in performance over existing approaches. ...

ABSOLUTELY AMAZING work: Behavioral Homophily in Social Media via Inverse Reinforcement Learning: A Reddit Case Study Yuan, Schneider and Rizoiu, WWW2025. Openreview. DOI: https://doi.org/10.1145/3696410.3714618. Arxiv. Online communities play a critical role in shaping societal discourse and influencing collective behavior in the real world. The tendency for people to connect with others who share similar characteristics and views, known as homophily, plays a key role in the formation of echo chambers which further amplify polarization and division. Existing works examining homophily in online communities traditionally infer it using content- or adjacency-based approaches, such as constructing explicit interaction networks or performing topic analysis. These methods fall short for platforms where interaction networks cannot be easily constructed and fail to capture the complex nature of user interactions across the platform. This work introduces a novel approach for quantifying user homophily. We first use an Inverse Reinforcement Learning (IRL) framework to infer users’ policies, then use these policies as a measure of behavioral homophily. We apply our method to Reddit, conducting a case study across 5.9 million interactions over six years, demonstrating how this approach uncovers distinct behavioral patterns and user roles that vary across different communities. We further validate our behavioral homophily measure against traditional content-based homophily, offering a powerful method for analyzing social media dynamics and their broader societal implications. We find, among others, that users can behave very similarly (high behavioral homophily) when discussing entirely different topics like soccer vs e-sports (low topical homophily), and that there is an entire class of users on Reddit whose purpose seems to be to disagree with others. ...

This is a reading note of an accepted oral paper at WWW2025: Welcome to the Dark Side — Analyzing the Revenue Flows of Fraud in the Online Ad Ecosystem By Emmanouil Papadogiannakis, Nicolas Kourtellis, Panagiotis Papadopoulos and Evangelos Markatos. Link: https://doi.org/10.1145/3696410.3714899 The online ad ecosystem is a mess—a densely tangled web of players with overlapping roles, and worse, a dictionary full of confusing terminologies. This paper takes a flashlight to one particularly shadowy corner (ad fraud?) beneath the glossy, “efficient” surface of digital advertising. ...

(Btw, all the WWW keynotes can be found here.) Foundation models are large, pre-trained neural network trained on broad, diverse time series data with the goal of supporting many downstream tasks (forecasting, anomaly detection, causal inference, generation) across different domains. It’s like GPT-4, but for time series analysis. 🔹 Core Capabilities of a Time Series Foundation Model: Prediction — short-term, long-term, multivariate Analysis — pattern discovery, representation learning Causal Inference — learning and modeling causal relationships over time Generation — synthetic time series for simulation or augmentation Cross-domain transfer — one model works across finance, medicine, climate, etc. To use a foundation model, there are mainly two ways: Prompt-based learning vs Fine-tuning. Prompting is fast (like how you ask ChatGPT questions). Fine-tuning can be costly but would be better for narrow, high-accuracy needs. ...

On April 21, 2025, Oh! The Federal Trade Commission filed a lawsuit today against Uber, alleging the rideshare and delivery company charged consumers for its Uber One subscription service without their consent, failed to deliver promised savings, and made it difficult for users to cancel the service despite its “cancel anytime” promises. Among the many allegations, the following may not guarantee a courtroom victory, but it stands out as the most egregious issue: ...

For the paper: Price Stability and Improved Buyer Utility with Presentation Design: A Theoretical Study of the Amazon Buy Box. Ophir Friedler, Hu Fu, Anna Karlin, Ariana Tang. Accepted at The Web Conference 2025. paper pdf poster ver 1.0 Here’s the poster for WWW2025 presentation: Thanks to my advisor Hu Fu for his ideas on this.

Dianping is the monopoly rating platform in China. They have huge market. With great power comes great responsibility. Dianping is going to be in so much trouble (antitrust and regulations :)) if they don’t take active actions to harness and make good use of their power. Here’s an interesting Chinese article from the pov of restaurants: 困在评分系统里的餐饮人 驳静 (https://mp.weixin.qq.com/s/ACtVi3hphs3ml5euBcKDlQ) Some translated snippets In big cities these days, people are used to being bribed [by restaurants] with freebies in exchange for a five-star review [on Dianping]. Two decades after its founding, Dianping’s rating mechanism quietly rewritten the ground rules of the restaurant business. ...