💪 ARM: Alignment-oriented retrieval method

If you find our code, or the paper helpful, please cite the paper

@article{chen2025can,
  title={Can we retrieve everything all at once? ARM: An alignment-oriented LLM-based retrieval method},
  author={Chen, Peter Baile and Zhang, Yi and Cafarella, Michael and Roth, Dan},
  journal={arXiv preprint arXiv:2501.18539},
  year={2025}
}

Overview

1. Information alignment: align_info.py
2. Structural alignment
   1. Expand the base search objects obtained from information alignment: align_structure_expand.py
   2. Filter the expanded search objects into drafts: align_structure_filter.py
3. Self-verification and aggregation
   1. Self-verification: verify.py
   2. Aggregation: aggregate.py

Setup

Download the data folder and put it in the root directory

The code can be executed in a pipelined fashion in the order mentioned in Overview, so

python align_info.py --> python align_structure_expand.py --> ...

Files with

parser = argparse.ArgumentParser()
parser.add_argument('-p', '--partition', type=int)
args = parser.parse_args()

support parallel execution. num_partitions specifies the the number of jobs executed in parallel. To run these files (e.g., align_info.py), you can use, for example,

python align_info.py -p 0 & python align_info.py -p 1 & ...

followed by merge(...) in align_info.py to merge the outputs

Contact

If you have any questions or feedback, please send an email to peterbc@mit.edu.