Generative protein language models are a natural way to design new proteins with desired functions.

Here, we characterize the radiology report as a distinct domain and introduce RaLEs, the Radiology Language Evaluations, as a benchmark for natural language understanding and generation in radiology.

In the field of natural language processing, the prevalent approach involves fine-tuning pretrained language models (PLMs) using local samples.

We then resample a dataset with these domain weights and train a larger, full-sized model.

These sentence pairs will serve as a means to evaluate machines' commonsense reasoning abilities under different extra-linguistic contexts.

Recently, commonsense reasoning in text generation has attracted much attention. Generative commonsense reasoning is the task that requires machines, given a group of keywords, to compose a single coherent sentence with commonsense plausibility. While existing datasets targeting generative commonsense reasoning focus on everyday scenarios, it is unclear how well machines reason under specific geographical and temporal contexts.

We then train and release a 10.7B-parameter