A distilled cross-encoder fine-tuned on the STS Benchmark for efficient semantic textual similarity scoring.