I went on a 2nd date last night and while there won’t be a 3rd (irreconcilable differences), she registered to same-day vote today!
Some of us will be in the same living room for an election watch party tho @andsunnysideup will be hard at work keeping @votewithsway running :)
Just voted for the first time in CA, and feeling insanely grateful to @andsunnysideup and the @votewithsway team for the work they’re doing to make voting better.
In undergrad, I took a class with Ed Glaeser during which he explained the paradox of voting — it’s largely irrational and more about “expressive utility.” This, amongst many other things, left me feeling quite jaded about politics and the possibility of effecting change.
Andrew’s ballot party last weekend has made me the most energized to vote since I was 18 and an avid local political junkie. When you vote with trusted blocs, you vote in community and with trust/information — and you know your collective group of as few as 10 votes could flip an election, which in turn drives real change.
Sway is using tech and the power of networks to turn expressive utility into real, accountable voting blocs. Shoutout Sway!
Go get your vote in today and use a Sway voter guide if you need help! 🤠🇺🇸
@jessiicawu Curious to see to what extent more data and empiricism will actually allow us to better model the world in those poorly modeled domains—not just the ones you mention of bio and cyber, but particularly ones in social science.
AI-generated summaries of history led to more liberal opinions compared to Wikipedia, while summaries by chatbots prompted to use a conservative framing produced more conservative opinions—but primarily among conservative readers. In PNAS Nexus: https://t.co/7h5RgyQ1Rj
Just landed in Suzhou for #EMNLP2025! 🛬🦀🍜
Excited to present our paper showing human preferences != true helpfulness on:
📅 November 6th, 10:30
📌 Hall C3, Poster Session 4
Also happy to chat about evals, human feedback, personalization, deep research, and vibe coding! 😁
🚨 New Paper! 🚨
We want ~helpful~ LLMs, and RLHF-ing them with user preferences and reward models will get us there, right?
WRONG! 🙅❌⛔️
Our #EMNLP2025 paper finds a major helpfulness-preferences gap: user/LLM judgments + agent simulations can totally miss what helps users
IF YOU'VE BEEN MEANING TO READ ABUNDANCE‼️
my friend matt and i are roadtripping to a forest airbnb north of SF to read Abundance over the wknd and discuss, if you are 21-24 and want to join us fill out this form
we have 5 spots (10/17-19)
https://t.co/KF6SSzEHOy
This was a really fun paper to put together with @rachelrudinger and @boydgraber, allowing me to vent many of my frustrations working with MCQA over the past year 😪🫡
Please check out the paper, we would love to hear your feedback! 📄👇
https://t.co/pqTmwxK5BN
Excited to share three papers at #EMNLP2024 on:
📊 Commonsense MCQA Quality Analysis
📚 NLP-Powered Flashcard Scheduling
🧑🎓 Aligning LLMs w/ Students for Mnemonic Generation
See you in Miami! 🎉🌴
🚨 New pre-print on AI and historical knowledge 🚨 https://t.co/OMrcpPWpRl
How does AI impact the way we learn?
We find that people shown AI-generated 🤖 summaries of historical events score better on knowledge questions than those shown human-written 🧑🦰💻 summaries
1/5
Clever SRS algorithm optimizer paper uses BERT-based similarity metric on item text to better approximate forgetting curves for new items (via @JarrettYe)
Paper URL: https://t.co/e5Fu0q0l0r
Via this very interesting SRS benchmarking post: https://t.co/DKduv9OyUz
📢📢Are you studying vocab for the #GRE/#GMAT and want to earn up to $50 while doing so? Come join our research study with KARL, a flashcard tool which now uses model-generated mnemonic devices to make vocab learning more engaging! 💡 📚 Details below👇
Can we enhance flashcard scheduling by harnessing the untapped textual content of flashcards?🤔
Excited to share our work on developing KARL, an NLP-powered adaptive learning tool! 🚀📚
Paper: https://t.co/r3p8SjqJFo
Live Demo: https://t.co/AOR0yVMiBt
Can we enhance flashcard scheduling by harnessing the untapped textual content of flashcards?🤔
Excited to share our work on developing KARL, an NLP-powered adaptive learning tool! 🚀📚
Paper: https://t.co/r3p8SjqJFo
Live Demo: https://t.co/AOR0yVMiBt
We believe machine learning can improve learning experiences with flashcards. Study #trivia with our app KAR³L (https://t.co/9nHKTbbtlF) to help us study this claim for research and receive compensation—link points to consent form and app! More details: https://t.co/kXkYHs1zcz
#RedefiningGeek#book#giveaway starts today! I'll be doing a few giveaways--the first is for #students. If you're a K-12 student, college student, or grad student, enter by May 24 by: 1) RTing this msg & 2) filling out this form https://t.co/6wC8WCLQ64 (Pic of @mrafalow bc 😍)
Reading @ruha9’s Race After Technology and decided to replicate the 2016 “three White/Black/Asian teenagers” search described in the book. White + Black mostly now return images from articles covering this incident. Asian still returns “scantily clad girls and women”…