Interactive interface: STaRK SKB Explorer

• [May 2024] 🔥 We have augmented our benchmark with three high-quality human-generated query datasets which are open to access. See more details in our updated arxiv!
• [May 11th 2024] We upgrade our Amazon knowledge base and uploaded datasets to huggingface. Now you can download the SKB data from our huggingface repo!
• [May 9th 2024] We release STaRK SKB Explorer on Hugging Face, an interactive interface for you to explore our knowledge bases! A demo video will be out soon.
• [May 7th 2024] We present STaRK in the 2024 Stanford Annual Affiliates Meeting and 2024 Stanford Data Science Conference.
• [May 5th 2024] STaRK was reported on Marketpost and 智源社区 BAAI. Thanks for writing about our work!
• [Apr 21st 2024] We release the STaRK benchmark.

STaRK is a large-scale semi-structure retrieval benchmark on Textual and Relational Knowledge Bases. Given a user query, the task is to extract nodes from the knowledge base that are relevant to the query.

• Novel Task: Recently, large language models have demonstrated significant potential on information retrieval tasks. Nevertheless, it remains an open question how effectively LLMs can handle the complex interplay between textual and relational requirements in queries.

• Large-scale and Diverse KBs: We provide three large-scale knowledge bases across three areas, which are constructed from public sources.

  

• Natural-sounding and Practical Queries: The queries in our benchmark are crafted to incorporate rich relational information and complex textual properties, and closely mirror questions in real-life scenarios, e.g., with flexible query formats and possibly with extra contexts.

  

Create a conda env with python 3.8 and install required packages in requirements.txt.

conda create -n stark python=3.8 
conda activate stark
pip install -r requirements.txt

from src.benchmarks.get_qa_dataset import get_qa_dataset
from src.benchmarks.get_semistruct import get_semistructured_data

dataset_name = 'amazon'

# Load the retrieval dataset
qa_dataset = get_qa_dataset(dataset_name)
idx_split = qa_dataset.get_idx_split()

# Load the knowledge base
kb = get_semistructured_data(dataset_name, download_processed=True)

Question answer pairs for the retrieval task are locally included in data/{dataset}/stark_qa. We provided official split in data/{dataset}/split.

There are two ways to load the knowledge base data:

• (Recommended) Instant downloading: The knowledge base data of all three benchmark will be automatically downloaded and loaded when setting download_processed=True.
• Process data from raw: We also provided all of our preprocessing code for transparency. Therefore, you can process the raw data from scratch via setting download_processed=False. In this case, STaRK-PrimeKG takes around 5 minutes to download and load the processed data. STaRK-Amazon and STaRK-MAG may takes around an hour to process from the raw data.

• Our evaluation requires embed the node documents into candidate_emb_dict.pt, which is a dictionary node_id -> torch.Tensor. Query embeddings will be automatically generated if not available. You can either run the following the python script to download query embeddings and document embeddings generated by text-embedding-ada-002. (We provide them so you can run on our benchmark right away.)

  python download_emb.py --dataset amazon --emb_dir emb/

  Or you can run the following code to generate the query or document embeddings by yourself. E.g.,

  python generate_emb.py --dataset amazon --mode query --emb_dir emb/ --emb_model text-embedding-ada-002

  • dataset: one of amazon, mag or primekg.
  • mode: the content to embed, one of query or doc (node documents).
  • emb_dir: the directory to store embeddings.
  • emb_model: the LLM name to generate embeddings, such as text-embedding-ada-002, text-embedding-3-large.
  • See generate_emb.py for other arguments.

• Run the python script for evaluation. E.g.,

  python eval.py --dataset amazon --model VSS --emb_dir emb/ --output_dir output/ --emb_model text-embedding-ada-002 --split test --save_pred 

  python eval.py --dataset amazon --model LLMReranker --emb_dir emb/ --output_dir output/  --emb_model text-embedding-ada-002 --split test --llm_model gpt-4-1106-preview --save_pred

  • dataset: the dataset to evaluate on, one of amazon, mag or primekg.
  • model: the model to be evaluated, one of VSS, MultiVSS, LLMReranker.
    • Please specify the name of embedding model with argument --emb_model.
    • If you are using LLMReranker, please specify API keys at config/openai_api_key.txt or config/claude_api_key.txt and the LLM name with argument --llm_model.
  • emb_dir: the directory to store embeddings.
  • split: the split to evaluate on, one of train, val, test, and human_generated_eval (to be evaluated on the human generated query dataset).
  • output_dir: the directory to store evaluation outputs.

Please consider citing our paper if you use our benchmark or code in your work:

@article{wu24stark,
    title        = {STaRK: Benchmarking LLM Retrieval on Textual and Relational Knowledge Bases},
    author       = {
        Shirley Wu and Shiyu Zhao and 
        Michihiro Yasunaga and Kexin Huang and 
        Kaidi Cao and Qian Huang and 
        Vassilis N. Ioannidis and Karthik Subbian and 
        James Zou and Jure Leskovec
    },
    eprinttype   = {arXiv},
    eprint       = {2404.13207},
  year           = {2024}
}