Hacker News new | ask | show | jobs
by famouswaffles 112 days ago
It's probably much worse than VLMs on the computer use benchmarks out there. A lot of those benchmarks would be very hard to complete without the intelligence that arises from text pretraining.