SwissGov-RSD: A Human-annotated, Cross-lingual Benchmark for Token-level Recognition of Semantic Differences Between Related Documents

This article introduces a new benchmark dataset, SwissGov-RSD, designed for evaluating models' ability to identify semantic differences at the token level across different languages. The focus is on cross-lingual understanding and the nuances of meaning within related documents. The use of human annotation suggests a focus on high-quality data for training and evaluation.

• •Introduces SwissGov-RSD, a new benchmark.
• •Focuses on token-level semantic difference recognition.
• •Employs human annotation for high-quality data.
• •Targets cross-lingual understanding.