This dataset is derived from the upstream MIMIR benchmark dataset (iamgroot42/mimir) and augments each example with multiple paraphrased variants produced by SAGE/SAGER pipelines using different LLM backends (DeepSeek-V3.2, Gemini-2.5-Flash, Grok-4.1-Fast).

Source subset: Wikipedia (EN)

Columns

member: member text nonmember: non-member text sage_deepseek: SAGE paraphrase using DeepSeek-V3.2 sage_gemini: SAGE paraphrase using Gemini-2.5-Flash sage_grok: SAGE… See the full description on the dataset page: https://huggingface.co/datasets/bilgehanertan/mimir-sage-paraphrased-wikipedia_en.