It’s been brewing for a while, and I’m excited to let you know that we have just launched the first continuous Entity Linking Benchmark for SNOMED CT, giving the clinical AI community a platform to test and compare how well different approaches map clinical free text to structured medical concepts.
Building on the success of our 2024 Entity Linking Challenge, we’ve used newly updated and annotated MIMIC-IV data to create something the community can keep coming back to.
Most AI benchmarks in healthcare test how well a model passes a medical exam or answers a patient’s questions. Very few test how well it handles real clinical notes. This benchmark changes that.
Go take a look and get submitting!