ManCAR

This repository provides a PyTorch reference implementation of the main models and training procedures described in our paper:

Kun Yang, Yuxuan Zhu, Yazhe Chen, Siyao Zheng, Bangyang Hong, Kangle Wu, Yabo Ni, Anxiang Zeng, Cong Fu, Hui Li. ManCAR: Manifold-Constrained Latent Reasoning with Adaptive Test-Time Computation for Sequential Recommendation.

Overview

we propose ManCAR, a principled framework that grounds reasoning within the topology of a global interaction graph. ManCAR constructs a local intent prior from the collaborative neighborhood of a user's recent actions, represented as a distribution over the item simplex. During training, the model progressively aligns its latent predictive distribution with this prior, forcing the reasoning trajectory to remain within the valid manifold. At test time, reasoning proceeds adaptively until the predictive distribution stabilizes, avoiding over-refinement.

Paper & Resources

• Hugging Face Papers: https://huggingface.co/papers/2602.20093
• arXiv: https://arxiv.org/abs/2602.20093
• Dataset (Hugging Face): https://huggingface.co/datasets/PIIR/ManCAR

Dataset process

you can download CDs dataset from Hugging Face

After downloading the dataset, you need put the dataset into dataset/processed/.

or use the following commands to process your datasets

1. Download the dataset from Amazon

2. python ./datasets/process_data.py

3. python ./datasets/item_csv.py

After processed, you need to put the processed dataset into dataset/processed/.

Requirements

torch==2.4.1

numpy

tqdm

Training

To run ManCAR, use the following command:

1. cd ManCAR
2. bash run.sh

Results

Acknowledgements

We greatly appreciate the official ReaRec repository. Our code is based on the ReaRec repository.

Citation

If you use this dataset, please cite:

@article{mancar2026,
  title={ManCAR: Manifold-Constrained Latent Reasoning with Adaptive Test-Time Computation for Sequential Recommendation},
  author={Kun Yang, Yuxuan Zhu, Yazhe Chen, Siyao Zheng, Bangyang Hong, Kangle Wu, Yabo Ni, Anxiang Zeng, Cong Fu, Hui Li},
  journal={arXiv preprint arXiv:2602.20093},
  year={2026}
}