1-Encyclopedia
Wikipedia: https://huggingface.co/datasets/wikimedia/wikipedia
(To-Do)The complete wiki data is classified to 4 domains, including: common sense, medicine related, law related, computer science+coding related
2-Specific Domain
PubMed: https://huggingface.co/datasets/qiaojin/PubMedQA
Legal Bench: https://huggingface.co/datasets/nguha/legalbench
Code: https://huggingface.co/datasets/code-search-net/code_search_net
Arxiv-CS: https://huggingface.co/datasets/arxiv-community/arxiv_dataset (filter with category tags)
1-Question answering
Natural Questions: https://huggingface.co/datasets/google-research-datasets/natural_questions
Trivia Questions: https://huggingface.co/datasets/mandarjoshi/trivia_qa
Squad: https://huggingface.co/datasets/rajpurkar/squad
Web Questions: https://huggingface.co/datasets/stanfordnlp/web_questions
2-Reasoning
MMLU: https://huggingface.co/datasets/cais/mmlu
Strategy QA: https://huggingface.co/datasets/wics/strategy-qa
HotPot QA: https://huggingface.co/datasets/hotpotqa/hotpot_qa
1.Qwen3-embedding-0.6B: https://huggingface.co/Qwen/Qwen3-Embedding-0.6B-GGUF
2.Qwen3-embedding-4B: https://huggingface.co/Qwen/Qwen3-Embedding-4B
3.Other..
1.Qwen3
2.LLama 3/4
1-Setup baseline
- 实验设定：1个client，包含wiki知识库（common sense）
- 测试数据：7个test task中的得分
2-Add wiki data
- 实验设定：4个client，分别包含wiki知识库（common sense, medicine related, law related, computer science+coding related）
- 测试数据：7个test task中的得分
3-Add domain data
- 实验设定：4个client，分别包含1个wiki知识库（common sense）和三个domain知识库（med, legal, CS+coding）
- 测试数据：7个test task中的得分
4-Add wiki data + domain data
- 实验设定：7个client，分别包含4个wiki知识库（common sense, medicine related, law related, computer science+coding related）和三个domain知识库（med, legal, CS+coding）
- 测试数据：7个test task中的得分
5-Performance with PoR, PoG algorithms
- 实验设定：4个client，分别包含1个wiki知识库（common sense）和三个domain知识库（med, legal, CS+coding）
- 测试数据： 1）对于query请求的响应时间（retrieval时间，generation时间） 2）添加隐私算法之后的响应时间（retrieval+PoR时间，generation+PoG时间） 3）PoR, PoG proof size 4）变换embedding向量长度之后的影响