Language Models are Surprisingly Fragile to Drug Names in Biomedical Benchmarks

Open in new window