
PhD student @PrincetonCS @PrincetonCITP | Computational Social Science, NLP, Social Compiting | Previous: @EPFL_en, @chub_fbk, @UniBo | 🦋 frasalvi
Joined November 2013
- Tweets213
- Following435
- Followers417
- Likes2.5K
Pinned Tweet
📢 I'm starting a Substack! It's called Resetmaxxing.
My Claude usage resets every Sunday morning, and I usually reach the weekend with some tokens to spare. So I’ve decided to point an AI agent at a hard, long-shot problem and let it run for a day or two. 🧵
Francesco Salvi retweeted
[Please reshare]
I’m recruiting PhD students to join my lab at @UW!
Our current focus is on understanding the effects of social media algorithms and finding ways to improve them, but we are increasingly interested in how our findings from social media apply to chatbots/LLMs.
📜 Resetmaxxing #001: Can an AI agent read the Voynich manuscript?
I let my agents spend 258 h trying to decode a mysterious medieval script that nobody has ever deciphered. It came back with an (overconfident) answer, but probably not the one you were hoping for.
Link below 👇
The write-ups are largely AI-generated from the agent’s notes, with my own take kept separate.
The launch post, with the whole story: resetmaxxing.substack.com/p/…
Got a problem you’d love to see an agent attempt? Send it my way!
1
28
Special thanks to @ricgallotti's BIG IF TRUE and @random_walker @sayashk's CRUX evals for the inspiration!
1
31
Francesco Salvi retweeted
What does ChatGPT tell folks about the midterm elections? It depends on who's asking!
We audited ChatGPT inside 160 real users' own accounts in the days before the midterms. With memory on, we find that answers shifted toward each user's own party!
elections.citp.princeton.edu…
1/4
What happens when two people ask ChatGPT the same political questions?
Well, it depends on what ChatGPT knows about them. 🧵
3/4
More broadly, we argue that audits need to incorporate real user histories alongside these more controlled approaches, or risk missing important aspects of how AI systems behave in practice.
1
3
54
4/4
📝 Read more: elections.citp.princeton.edu…
Stay tuned for more to come 🔜
2
46
Francesco Salvi retweeted
What happens when a Wikipedia language edition starts using LLMs?
Unlike in other languages, Korean Wikipedia embraced LLMs through WikiVault: only 0.1% of editors use it weekly, but it accounts for ~30% of added bytes!
New preprint led by @song_inhwa_ (who's at @ACM_CSCW)
🧵
Francesco Salvi retweeted
🚨SciConBench accepted at #NeurIPS2026!
Paper: arxiv.org/abs/2606.11337
Can AI *synthesize* scientific conclusions in health? We find that, on medical conclusions, we are far from it!
Check out our live dashboard: sciconbench.cs.princeton.edu…
w/@hayounggjung and @korolova
Headed to Vermont for IC2S2!
I will be giving two parallel talks:
🎤 Wed 29, 10:45-12:15 (AI in Education). Ongoing work w/ @WikiEducation on the impact of LLMs in Wikipedia.
🎤 Wed 29, 2:15-3:45 (HCI and practice). LLMs persuasiveness in commercial settings.
Francesco Salvi retweeted
Thrilled to share that I am joining UC Berkeley as an Assistant Professor in the School of Information!
I start in Fall 2027, and I am recruiting PhD students this cycle. List me in your application if you're interested in frontier AI evaluation, AI policy, and AI's impacts on institutions such as science, law, and medicine.
I'm especially keen to work with students interested not just in high-quality research, but also in communicating it with a broad audience such as by public writing and policy impact. Fill out the form in the next tweet to indicate your interest.
As for this coming year, I'm moving to Berkeley this fall to start something new with @RishiBommasani and @random_walker. We'll have much more to share soon.
Francesco Salvi retweeted
Francesco Salvi retweeted
New preprint!
We introduce a new benchmark, SciConBench, with 9.11k scientific questions derived from Cochrane Systematic Reviews.
We find evidence that frontier AI agents **cannot** synthesize scientific conclusions well.
A thread 🧵
w/ @hayounggjung, @korolova & others















