angela zhou

I just wish I could make the equivalency network of ideas and tools and missions real and interpersonal and convivial

i bought this gorgeous limited edition sundae school venice "beach" fleece only to realize it's too heavy to wear in LA like, ever ever .........

im in the uk in my interdisciplinary black hole just trying to work a lot and grind out papers - v appreciative it's cold enough here for me to live in my graphic fleece

i submitted a revision last night and bought a graphic fleece in celebration and the shopkeep had to point out that it looks just like the one i was wearing at the time fml
me: Leave me alone in my pile of cozy extremely graphic fleeces

Figure 2: Fields vary widely in their consensus around top-5 venues and the general alignment of their prefer- ences. Whether one uses a field-level ranking constructed from the pairwise comparisons of all respondents in a field to predict an individual’s pairwise comparisons (horizontal axis) or assesses the degree to which individuals in a field agree on a top-5 (vertical axis), some fields (e.g. Economics) demonstrate a much more overlapping and organized set of preferences than others (e.g. Computer science).

Academic fields differ in their degree of consensus about what "better" publication venues are — here shown both for consensus about the "top 5" and for pairwise choices.

Hello #econsky :)

arxiv.org/abs/2603.00807
@jugander.bsky.social @danlarremore.bsky.social @aaronclauset.bsky.social

I often write "sermons" to our applied math PhD students as part of my Friday roundup email. This week, I wrote about the use of AI (and specifically LLMs) in applied math research:
www.cs.cornell.edu/~bindel/serm...

feeling like an interdisciplinary leper living somewhere in the black hole of "you're more of a x than a y" (says y) and a "you're more of a y than an x" (says x) ☠️

Oh very cool, thank you!!

Benchmarks Saturate When The Model Gets Smarter Than The Judge arxiv.org/pdf/2601.19532 may be worth a look? Recent preprint which looks specifically at LLM-as-judge failures for the Omni-MATH dataset

We actually found (following some earlier work) that the pairwise/pointwise difference does not actually matter that much, and the intuition that pairwise judgments are better because they have a reference point may not hold true

aclanthology.org/2025.finding...
aclanthology.org/2025.emnlp-m...

amazing, thanks!

gsood.com/research/pap...

This may be helpful on both fronts! + refs in the lit review

Dear hivemind,
Do you have any favorite references on failures of LLM-as-a-judge (and universally enforceable validity checks? besides some kind of cohen's kappa on a small subset)

“In the last year, it’s gotten a lot worse” A Qualitative Investigation of Barriers to Disability Benefits in 2025 - DREDF This report is based on interviews with 52 benefits professionals serving over 8,000 disability claimants nationwide. The independent research finds that recent Social Security Administration administ...

In this new report, @ksavin.bsky.social, and @calliefreitag.bsky.social and I explain how Trump 2.0 and DOGE built up new barriers to the already-difficult process of getting disability benefits.

Let's talk about what we found, and what we recommend to address the issues. 🧵

they added an AI admin to review my concur request 😁

Difficulties accessing disability benefits: “Getting stuck in a loop.” We identified five major areas... (a) changes to the phone system, (b) field offices’ appointment and walk-in policies, (c) staffing cuts and reassignments, (d) increased processing times and denials, ..

Sorry to hear this Alexandra :( Big fan of your work and the whole FATE Montreal team - this is a huge loss !!!

It's been a journey of nearly 3 years, but I'm very excited to announce the CNTR AISLE Portal! 🚀 cntr-aisle.org It’s a new way to review and evaluate the 1,000+ AI bills introduced in the U.S. over the last three years. Check out the Bill Library and our Profiles#AIPolicy #OpenData

Mamdani’s New Social Services Chief Worked to Reduce Street Homelessness

www.nytimes.com/2026/02/25/n...

Big news - allegheny county's erin dalton tapped for ny social services chief role

A picture of Joe Halpern smiling in green shirt in front of a blue background.

Today arXiv remembers our colleague Joe Halpern, who was instrumental in founding arXiv's CS section.

Joe's passions ranged far & wide and we're lucky that arXiv was one of them. Joe, thank you for giving so much to arXiv - you are missed.

blog.arxiv.org/2026/02/27/remembering-joe-halpern

Democrats to introduce bill aimed at resurrecting IRS Direct File Democrats want to bring back Direct File.

You can kill the product but not the idea - today, 160 Congressional co-sponsors introduced a bill to bring back the IRS’ much beloved free tax filing product, Direct File.

www.theverge.com/policy/88522...

Acemoglu, Autor & Johnson sketching out a pro-worker AI agenda. Worth your time.

old-school, but Sophie Calle worked a lot with surveillance (La Filature / The Shadow - hired a PI) hyperallergic.com/following-so... . could be a nice historical touchpoint?

"On a side note, Alexa Plus on the Echo Show has a new interface that pops up when motion sensing detects someone nearby. It includes chatty messages like "Reminder: I'm always in your corner" and "Ready to dive into whatever's on your mind," ... "
Who signed off on this
www.cnet.com/tech/service...

Big Game Commercial | Alexaaaa+ (Extended Cut) YouTube video by Amazon Alexa

catching up on superbowl ads. www.youtube.com/watch?v=sCyi...
the nerds are out of touch 🫣
just show me ai that will do my laundry 🫣

Took a break from my 25+ pg and growing response to referees to yak shave a rebuttal down to 5000 chars 🐑✂️🙏💆‍♀️

"I seem unreliable, but I'm actually very reliable. You just have to trust the process." Alysa Liu, U.S. Olympic figure skater

When I’m two weeks behind on my grading.

the past couple of years I've spent a lot of time digging into applied collaborations/projects. I feel like a lot of theory-first people write this off as not a valuable mode of thinking, but the truth is I find it useful for me to believe in something, and it's certainly not techco's narratives atm

wish there could be like the equivalent of knitting or crafting circles where reasonable folks who are reasonably skeptical and tired of ai hype can get together and build little better futures together