agentlans/markov-slop
收藏官方服务:
资源简介:
Markov Slop是一个通过可变阶马尔可夫链语言模型生成的合成文本数据集。它包含长文本块,这些文本块表现出局部句法正确性,但缺乏全局语义连贯性。该数据集专为NLP基准测试、基线语言建模、困惑度分析以及AI安全/内容检测研究而构建。
Markov Slop is a synthetic text dataset generated via a variable-order Markov chain language model. It contains long-form text blocks that exhibit local syntactic correctness but lack global semantic coherence. This dataset is purpose-built for NLP benchmarking, baseline language modeling, perplexity analysis, and AI safety/content detection research.
提供机构:
agentlans


