Tag
This paper investigates how the size of text chunks (sentences, paragraphs, chapters) in Retrieval-Augmented Generation systems affects generation quality, retrieval precision, and computational efficiency, using identical textbooks segmented at different granularities.
CobSeg is a multi-branch architecture for dialogue topic segmentation that models coherence boundaries and improves lexical transition detection, outperforming prior non-LLM approaches on multiple benchmarks.