RT @Get_Writer: Today is the day you begin delegating complex knowledge work to your own digital teammate. We’re beta launching our new au….

RT @WiAIR_podcast: Watch the full episode 👇.🎥 Youtube: 🎧 Spotify: 🎧 Apple: .

RT @WiAIR_podcast: What if you could automatically detect and remove mislabeled image-caption pairs?🤯.🎙️ In our new WiAIR episode, we sat d….

Huge thanks to my co-authors for their collaboration and guidance throughout the project!!.

Unlike prior methods that operate on expected amortized attention for each individual, we also define new divergence-based measures for attention distribution-based fairness in ranking (DistFaiR), with useful properties.

We observe that metrics for fair ranking are indeed sensitive to query polarity in practice, and an adjustment to fair ranking metrics to account for query properties such as polarity.

We find that prior research in amortized fair ranking ignores critical information about queries, and treats all queries as equivalent ignoring query properties in fairness measurement. This increases the risk of *fairwashing*.

Machine learning-driven rankings, where individuals are ranked in response to a query, mediate exposure in a variety of safety-critical settings. In this work, we examine amortized fair ranking -- where relevance and attention are cumulated over a sequence of user queries.

Excited to share our #WWW2025 paper ( with @kaiwang_gua , @walesalaudeen96 , .@asia_biega, and @MarzyehGhassemi!. Paper highlights in thread below:

RT @WiAIR_podcast: Machine learning has advanced rapidly, but its real-world impact in healthcare remains limited. Why?.🎙️ In our NEW EPISO….

RT @WiAIR_podcast: 🚀 Our new episode is LIVE! 🎙️.In Episode 3, we talk with @AparnaBee about:. 🏥⚠️ Unique challenges of applying AI in medi….

Presenting posters today at the AFME workshop (West 111-112), and the RBFM workshop (West 217-219)! Drop by to chat more about automatic data filtering and fair ranking! #NeurIPS2024.

RT @kaiwang_gua: I am finally done with the semester and on my way to catch the end of #NeurIPS2024 tomorrow. Come to our poster and talk t….

Some links:.Website: Work I'm presenting at NeurIPS:

Feel free to drop by and say Hi or reach out if I'd be a good fit for any relevant postdoctoral or research positions! . Also always happy to chat with students exploring the idea of pursuing a PhD -- happy to help if I can!. (5/n).

This is joint work with @kaiwang_gua , @walesalaudeen96 , @asia_biega , and @MarzyehGhassemi . Presenting at the poster session in the AFME workshop, December 14th. (4/n).

At the Algorithmic Fairness through the Lens of Metrics and Evaluation Workshop, presenting recent work on fair ranking. We propose distribution-aware metrics for measuring amortized fairness in rankings, and show the impact of query polarity in fair ranking. (3/n).

This is work co-led with Haoran Zhang (, and joint work with Nassim Oufattole, @HyewonMandyJ , Yan Wu, @JiachengZhu_ML , and @MarzyehGhassemi . 6:30-8pm PST, December 10th and 09:50 – 10:45am PST, December 14th. (2/n).

At Responsibly Building Foundation Models and Women in Machine Learning Workshops, presenting recent work on automatic label error detection using VLMs. TL;DR: We propose a method to identify label errors in large image captions datasets using VLMs. (1/n).