README.md

metadata

task_categories:
  - image-text-to-text
  - visual-question-answering
language:
  - en
tags:
  - visual-question-answering
  - multimodal
  - reinforcement-learning
  - visual-reasoning
  - spatial-reasoning
  - transit-maps

ReasonMap-Plus Dataset

The ReasonMap-Plus dataset is an extended dataset introduced in the paper RewardMap: Tackling Sparse Rewards in Fine-grained Visual Reasoning via Multi-Stage Reinforcement Learning. It addresses the challenge of sparse rewards in fine-grained visual reasoning for multimodal large language models (MLLMs), particularly in structured and information-rich settings like transit maps. ReasonMap-Plus achieves this by introducing dense reward signals through Visual Question Answering (VQA) tasks, which enables effective cold-start training of fine-grained visual understanding skills. This repository contains the ReasonMap-Plus data for evaluation and ReasonMap-Train for RewardMap training.

• Paper: RewardMap: Tackling Sparse Rewards in Fine-grained Visual Reasoning via Multi-Stage Reinforcement Learning
• Project Page: https://fscdc.github.io/RewardMap
• Code: https://github.com/fscdc/RewardMap

Sample Usage

To get started with the ReasonMap-Plus dataset, follow these steps to install dependencies, download the data, and prepare it for training.

1. Install dependencies

If you face any issues with the installation, please feel free to open an issue on the GitHub repository.

pip install -r requirements.txt

2. Download the dataset

You can download ReasonMap-Plus (for evaluation) and ReasonMap-Train (for RewardMap training) from Hugging Face or by running the following command:

python utils/download_dataset.py

Then, put the data under the folder data.

3. Prepare data for Supervised Fine-Tuning (SFT)

If you plan to use tools like LLaMA-Factory for SFT training, first prepare the datasets by running the following command:

python utils/prepare_data_for_sft.py --dataset_dir path/to/your_data

4. Data Format Example

Your data will be transferred into a format similar to this for SFT:

  {
    "conversations": [
      {
        "from": "human",
        "value": "<image> Please solve the multiple choice problem and put your answer (one of ABCD) in one \"\\boxed{}\". According to the subway map, how many intermediate stops are there between Danube Station and lbn Battuta Station (except for this two stops)? \
A) 8 \
B) 1 \
C) 25 \
D) 12 \
"
      },
      {
        "from": "gpt",
        "value": "B"
      }
    ],
    "images": [
      "./maps/united_arab_emirates/dubai.png"
    ]
  },

5. Training Example

You can train the RewardMap model using the provided scripts:

# RewardMap training
bash scripts/reward_map.sh

Citation

If you find this paper useful in your research, please consider citing our paper:

@article{feng2025rewardmap,
  title={RewardMap: Tackling Sparse Rewards in Fine-grained Visual Reasoning via Multi-Stage Reinforcement Learning},
  author={Feng, Sicheng and Tuo, Kaiwen and Wang, Song and Kong, Lingdong and Zhu, Jianke and Wang, Huan},
  journal={arXiv preprint arXiv:2510.02240},
  year={2025}
}