Our latest news, articles, and announcements
Finding the top-K methods, without wasting comparisons
If all you need are the winners, ranking the rest is wasted budget. Top-K spends every comparison separating the leaders and found the true top three in ten runs out of ten.
January 19, 2026
Evaluating interactive avatars
Interactive avatars are virtual personas that interact with users in real time through voice, gestures, and visual animations. As interactive avatars become more widely deployed in high-stakes domains such as education, customer support, and healthca
December 13, 2025
Measuring the progress of learned image compression
We compared the results of the recent Challenge on Learned Image Compression with last year's results, and found no significant improvement in perceptual quality. Does this mean that progress has stalled? Let's take a closer look.
November 21, 2025
Simplifying perceptual quality evaluation
Today, we are excited to launch Mabyduck, a self-serve platform that makes it incredibly easy to perform subjective quality evaluation experiments.