Developed a targeted feature enhancement for the rwth-i6/i6_core repository, focusing on improving Word Error Rate (WER) evaluation in speech recognition workflows. Introduced two new Python-based job classes that integrate with existing command line tools and data processing pipelines. The RescoreLatticeCacheJob enables flexible rescoring of lattice caches using configurable rescorer types and parameters, while the IntersectStmCtmJob computes intersections between CTM and STM files to ensure accurate WER evaluation on filtered corpora with matched recordings. This work deepened the repository’s evaluation capabilities, supporting more reliable scoring, faster iteration, and greater flexibility in configuring speech recognition experiments and analyses.