@iroasmas Catholic drip is just late Roman Republican wardrobe so this is actually just another way for men to never stop thinking about the Roman Empire
@Jabaluck@erinhengel And when Erin ends up being right and this whole group laughably wrong in a decade, to what will you attribute your overconfidence here?
Unified multimodal models can generate text and images, but can they truly reason across modalities? 🎨
Introducing ROVER, the first benchmark that evaluates reciprocal cross-modal reasoning in unified models, the next frontier of omnimodal intelligence.
🌐 Project: https://t.co/qA5EPaK5s7
📄 Paper: https://t.co/UjLGs3ZFel
📂 Benchmark: https://t.co/2oyk8SyYOo
@rohanpaul_ai A well-known principle of automation is that you should solve a problem yourself manually before automating the solution. In keeping with this principle, I think that we need to go through a stage of cognitive augmentation:
https://t.co/uIAEsMbx5B
New paper 📜: Tiny Recursion Model (TRM) is a recursive reasoning approach with a tiny 7M parameters neural network that obtains 45% on ARC-AGI-1 and 8% on ARC-AGI-2, beating most LLMs.
Blog: https://t.co/w5ZDsHDDPE
Code: https://t.co/7UgKuD9Yll
Paper: https://t.co/3m8ANhNMiw
arc-agi 2 is basically being solved
poetiq has reached 75% on arc-agi-2 using gpt-5.2 xhigh, averaging around $8 per task
that's well above the reported human average of 60%.
if you're confused, they're not training a new model from scratch
just wrapping existing models in a partly proprietary system to solve challenges like the arc prize
FYI, for anyone who knows me from 2022-2024, I am here primarily (perhaps only) to keep up with the latest AI research. If you want "old normality" style discourse, my main haunt these days is Substack: https://t.co/fxxkar58hS