README.md

metadata

language:
  - en
license: cc-by-sa-4.0
task_categories:
  - text-generation
tags:
  - recommendation-system
  - user-simulation
dataset_info:
  features:
    - name: text
      struct:
        - name: action_list
          sequence: string
        - name: history
          dtype: string
        - name: profile
          dtype: string
    - name: user_id
      dtype: string
    - name: item_id
      dtype: string
    - name: timestamp
      dtype: timestamp[ns]
    - name: item_pos
      dtype: int64
    - name: choice_cnt
      dtype: int64
    - name: dataset
      dtype: string
    - name: impression_list
      sequence: string
  splits:
    - name: test
      num_bytes: 14699116
      num_examples: 6400
  download_size: 5555631
  dataset_size: 14699116
configs:
  - config_name: default
    data_files:
      - split: test
        path: data/test-*

UserMirrorer-eval

This is the evaluation set of UserMirrorer, presented in the paper Mirroring Users: Towards Building Preference-aligned User Simulator with User Feedback in Recommendation.

Code: Joinn99/UserMirrorer

Notice

In the UserMirrorer dataset, the raw data from MIND and MovieLens-1M datasets are distributed under restrictive licenses and cannot be included directly.

Therefore, we provide a comprehensive, step-by-step pipeline to load the original archives, execute all necessary preprocessing operations, and assemble the final UserMirrorer training and test splits.

Click here to run the script notebook on Google Colab to get the full dataset. Also, you can download it and run it locally.

Evaluation Usage

To run the evaluation, you can execute the following command provided in the official repository:

python usermirrorer/run_eval.py \
    --project_path <YOUR_WORKING_DIR> \     # The path to your working directory
    --model_path <MODEL_PATH> \             # The path to the model
    --input_file <INPUT_FILE> \             # The path to the input file
    --output_file <OUTPUT_FILE> \           # The path to the output file
    --mode <MODE> \                         # The mode of the evaluation
    --repeat_times <REPEAT_TIMES> \         # The number of sampling times

Citation

@misc{wei2025mirroringusersbuildingpreferencealigned,
      title={Mirroring Users: Towards Building Preference-aligned User Simulator with User Feedback in Recommendation}, 
      author={Tianjun Wei and Huizhong Guo and Yingpeng Du and Zhu Sun and Huang Chen and Dongxia Wang and Jie Zhang},
      year={2025},
      eprint={2508.18142},
      archivePrefix={arXiv},
      primaryClass={cs.HC},
      url={https://arxiv.org/abs/2508.18142}, 
}