Reproduce “CNNSum: Exploring Long-Context Summarization with Large Language Models in Chinese Novels” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Lingxiao Wei, He Yan, Xiangju Lu, Junmin Zhu, Jun Wang, Wei Zhang
The authors introduce CNNSum, a benchmark of 695 Chinese novel samples with human-written summaries at lengths from 16k to 128k tokens. They fine-tune Qwen1.5-7B on short-context data. PI scale 2 scores 15.1 on CLongEval-LStSum Large and 7.4 on CNNSum-3XL; scale 4 scores 14.1 and 12.7, respectively. The ROUGE-L results show a ranking reversal for this setup and test pair, not human summary quality.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.