One paper by DFKI-NLP authors accepted to COLING 2025

One paper originating from research in the TRAILS project has been accepted to the Main Track at the 31st International Conference on Computational Linguistics 2025 (COLING 2025). In the paper, we present CROSS-REFINE, a generator-critic framework that enhances natural language explanations by refining initial outputs using feedback from a second LLM, outperforming SELF-REFINE and working effectively even with less powerful models.