- Adds the ability to generate sentence-level explanations with MSP instead of explanations of fixed-length text blocks.
- Speeds up MSP by collecting masked sequences into batches before running inference to measure differences in predicted probabilities.