ImplexConv

We design a large-scale multi-session conversation dataset to study implicit reasoning in personalized conversations and a hierarchical tree framework for for efficient, level-based retrieval.

Installation

conda create -n ImplexConv python=3.9
conda activate ImplexConv
python -m pip install -r requirements.txt

If you need to use OpenAI APIs, you will need to obtain an API key here.

export OPENAI_API_KEY=[your OpenAI API Key]

Dataset

All datasets referenced in the paper are available on HuggingFace.

Usage

1. Create conversation summarization and facts:

python fact_sum_batch.py \
    --home_dir ./datasets \
    --dataset_name opposed_reasoning \
    --model_type gpt-4o-mini \
    --output_file summarized_opposed_facts.json

3. Generate the response and retrieved content:

python fact_topic_reasoning.py \
    --home_dir ./datasets \
    --dataset_name opposed_reasoning \
    --model_type gpt-4o-mini \
    --summy_info summarized_opposed_facts.json \
    --output_response_file opposed_response.json \
    --output_retrieve_file opposed_retrieved_text.json

Citation

If you find the work useful, please cite:

@article{li2025toward,
  title={Toward Multi-Session Personalized Conversation: A Large-Scale Dataset and Hierarchical Tree Framework for Implicit Reasoning},
  author={Li, Xintong and Bantupalli, Jalend and Dharmani, Ria and Zhang, Yuwei and Shang, Jingbo},
  journal={arXiv preprint arXiv:2503.07018},
  year={2025}
}