@ShangyinT You have to manually check it. But the good news is that definitions and theorems are a small part of the formalization. The bulk is proofs.
English
Aws Albarghouthi
2.1K posts
@awsTO I’m always curious what’s the standard for ensuring the formalization matches the actual theorem?
English
It took me a few hours to write a Lean formalization of the theorems and proofs in a recent paper of ours. I used Fable + 5.6 Sol. I found the exercise shockingly eye opening. I don't think we should publish papers without Lean proofs at this point.
English
My only interventions were aesthetic: "this definition is too general" and "use mathlib here instead"
English
Aws Albarghouthi retweetledi
Is there anything computers can't do? Does Alan Turing's revolutionary work remain relevant in a world with generative AI and quantum computers? Announcing a new mini-series, for a general audience, on Computation and Its Limits.
This is one of several short courses on the new Ergo platform, "a nonprofit that publishes long-form philosophical lectures online, without ads or paywalls."
Get started at ergo.org/courses/comput…
English
in 1948, Turing introduced "unorganized machines", machines that learn, which we now call neural nets.
in 1949, in "Checking a large routing", Turing introduced formal verification (Floyd-Hoare triples).
Today, neural nets do formal verification. Pretty wild.
English
Aws Albarghouthi retweetledi
Excited to share that Train-to-test (T^2) scaling was accepted to COLM! 🌉
We show that when you factor test-time scaling into pretraining scaling, extreme overtraining becomes compute optimal.
Check out our paper below! 👇
Nicholas Roberts@nick11roberts
That new LFM2.5-350M is super overtrained, right? And everyone was shocked about how far they pushed it? As it turns out, we have a brand new scaling law for that! 🧵 [1/n]
English
excellent work from @GOrlanski and company!
Henry Kiss Ehrenberg@henryehrenberg
We expect agents to act like senior engineers, but most benchmarks still evaluate them like interns. Excited to introduce Senior SWE-Bench, an open-source and @harborframework-native benchmark that assesses agents as senior engineers on long-horizon tasks with realistically under-specified instructions. We expect agents to build real features going on just a quick Slack message, nothing like the super technical instructions most benchmarks provide. Senior SWE-Bench fixes that. Claude Opus 4.8 is the current leader at 24% high quality solves, but it took 117K tokens on average to get there. Claude Sonnet 5 looked like it was going to swoop in for the top spot, but we found it cheated on 26% of trials.
English
Aws Albarghouthi retweetledi
On the pod: "Constrained Adaptive Rejection Sampling" with @ucsd_cse professor @lorisdanto.
Hear how symbolic AI experts have navigated the LLM era and why the future of AI code generation depends on program synthesis and formal methods.
English
@moarbugs @AntithesisHQ Congratulations, Rohan! Enjoy the sabbatical!
English
Effective today, I've been promoted to Associate Professor at CMU. Thanks to my amazing students, colleagues, and peers for the support!
Now time for a sabbatical :-)
I'm incredibly excited to have joined @AntithesisHQ at the absolute forefront of autonomous software testing!!
English
Aws Albarghouthi retweetledi
Aws Albarghouthi retweetledi
The Khatib family was killed not only because of Iranian missiles, but also because the Israeli government’s neglect and failure to protect its Palestinian citizens, systematically refusing to build shelters in their towns.
Ihab Hassan@IhabHassane
Tonight, Iranian missiles killed four Palestinian girls from the same family in the town of Tamra, northern Israel. -Manar Al-Fakhri Khatib -Manar Al-Qassem Abu Al-Heija / Khatib -Shatha Khatib -Hala Khatib May their memory be a blessing.
English