+ 
+
+ Models that keep learning. +
+ + + +--- + unify is a research lab working on continual learning: systems that improve from a stream of experience without forgetting what they know. -[unify.ai](https://unify.ai) +Today's models are trained once and frozen. Everything they learn after deployment is thrown away. We study how a single system can keep learning, from new data, new tasks, and its own mistakes, while holding on to what it already knows. That means consolidation, plasticity, and the benchmarks that measure them honestly. + +The code here is what we build along the way: training and evaluation code for our papers, the benchmarks we use, and the tooling around them. + +### Learn more + +- **[unify.ai](https://unify.ai)**: what we are working on and who we are. +- **[Discord](https://discord.com/invite/sXyFF8tDtm)**: follow along and ask questions. diff --git a/public_images/unify-banner-green.png b/public_images/unify-banner-green.png new file mode 100644 index 0000000..066d026 Binary files /dev/null and b/public_images/unify-banner-green.png differ diff --git a/public_images/unify-banner-light.png b/public_images/unify-banner-light.png new file mode 100644 index 0000000..671ad51 Binary files /dev/null and b/public_images/unify-banner-light.png differ