I vibecoded a formal verification tool for neural networks from scratch using two months and $3,000. It just placed 2nd of 13 tools in VNN-COMP 2026.
Read more: https://t.co/Qr5F5EF8PS
#VNNCOMP@confCAV@FLoC2026#ClaudeCode@claudeai
@sbucompsc@stonybrooku@OpenAI Also had some sweet graphics for the hackathon from DALL-E 2: "Lecture with hundreds of students being taught by AI hologram professor." (left); "Wolf mascot typing on a laptop with a rainbow explosion of code coming out of the screen." (right)
We ran our first generative AI hackathon at Stony Brook University a few weeks ago. Students competed using generative AI like ChatGPT to make software complex programs.
https://t.co/8hjhpczQbj
@sbucompsc@stonybrooku@OpenAI#ChatGPT
Stony Brook University's AI Institute recently had an article featuring some of our recent work on verifying neural networks to avoid air-to-air collisions.
https://t.co/VjvjBD1tPE
@stonybrooku@AI_SBU@sbucompsc
@moarbugs For you CPS module, I would recommend our NFM '22 paper on finding rare errors in closed-loop systems with neural network controllers: https://t.co/wbyiiD3Fqk
Last week I was awarded the Research Excellence Award from Stony Brook University's Computer Science department! This is probably a good time to thank my research group (who did most of the work).
@sbucompsc@CEASSBU@StonyBrookU@SUNY
The third international competition on neural network verification (VNN-COMP) finished last weekend! Congratulations to the ฮฑ,ฮฒ-CROWN team (@huan_zhang12@KaidiXu1@ShiqiWang10) and all the participants! Full slides with more results: https://t.co/Sk8I5DtEct
Our SBU group placed first in the #GRAIC autonomous racing competition at #CPSIoTWeek (solo race category) which uses CARLA for simulation and has unknown obstacles on the track for you to avoid. Congratulations Abhinav Chawla and Adithya Guduru!
@SUNY@StonyBrookU@sbucompsc
Should conferences strive for a low acceptance rate? The most recent ACM SIGBED blog post argues excessive rejection may be hurting the CS research community.
New this year AAAI-23 will feature a track on Safe and Robust Artificial Intelligence, focusing on the theory and practice of safety and robustness in AI-based systems.ย Full details: https://t.co/579SL2oWj0
#AAAI23#SRAI#AI#artificalintelligence
A few weeks ago we had two papers at NASA Formal Methods #NFM2022 in @NASAJPL/@caltech: (1) exhaustive safety verification for closed-loop AI systems (https://t.co/wbyiiD3Fqk) and (2) shielding AI systems (Simplex) without a trusted backup controller (https://t.co/XDemawEY5h).