--- base_model: - Qwen/Qwen2.5-VL-7B-Instruct language: - en library_name: transformers license: apache-2.0 pipeline_tag: image-text-to-text tags: - multi-modal - large-language-model ---

MMR1: Advancing the Frontiers of Multimodal Reasoning

If you like our project, please give us a star ⭐ on Github to support us. 🙏🙏
## 📰 News * **[2025.03.11]** 🔥🔥 Release MMR1-Math-v0, achieving SOTA with only 6k data! ## Links Code: https://github.com/LengSicong/MMR1 This model was presented in the paper [LMM-R1: Empowering 3B LMMs with Strong Reasoning Abilities Through Two-Stage Rule-Based RL](https://arxiv.org/abs/2503.07536). Code can be found at https://github.com/LengSicong/MMR1 ## Model Description MMR1-Math-v0-7B is a Large Multimodal Model specialized in mathematical tasks. Remarkably, MMR1-Math-v0-7B achieves state-of-the-art performance among open-source 7B multimodal models, competing effectively even against proprietary models with significantly larger parameter sizes—all trained using only 6k carefully curated data instances. ### Key Highlights: - **SOTA Performance**: Sets a new **state-of-the-art** benchmark on math-related multimodal tasks among open-source 7B models. - **Minimal Training Data**: Remarkably achieves top-tier performance with just **6k** high-quality samples from **public training datasets**. - **Efficient Training with GRPO**: 6 hours of RL training with 64 H100s for 15 epochs. - **Public and High-Quality Data**: Publicly sourced datasets, rigorously filtered and balanced across both difficulty and mathematical problem types. - **Balanced Data Strategy**: Uniform sampling of data based on both task difficulty (filtering out overly simple problems) and mathematical reasoning diversity. ## Evaluation Results We evaluated our model using [VLMEvalKit](https://github.com/open-compass/VLMEvalKit/tree/main) on four mathematical reasoning benchmarks: MathVista_MINI, MathVision, LogicVista, and MathVerse_MINI. We also include results on the MathVerse_MINI_Vision_Only_cot (MathVerse_V) subset to maintain consistency with the VLMEvalKit leaderboard. The table below compares our model's performance against various open-source and proprietary models. | Model | size | MathVista | MathVision | LogicVista | MathVerse | MathVerse_V | |-------|:----:|:--------------:|:----------:|:----------:|:--------------:|:-------------------:| | **Close-sourced** | | | | | | | | [GPT-4o 1120](https://openai.com/index/gpt-4o-system-card/) | - | 60.0 | 31.2 | 52.8 | 40.6 | - | | [Gemini-2.0-flash](https://deepmind.google/technologies/gemini/flash/) | - | 70.4 | 43.6 | 52.3 | 47.8 | - | | [Claude3.7-Sonnet](https://www.anthropic.com/news/claude-3-7-sonnet) | - | 66.8 | 41.9 | 58.2 | 46.7 | - | | **R1-related** | | | | | | | | [LLaVA-CoT](https://github.com/PKU-YuanGroup/LLaVA-CoT) | 11B | 52.5 | 19.9 | 39.6 | 22.6 | - | | [Open-R1-Multimodal](https://github.com/EvolvingLMMs-Lab/open-r1-multimodal) | 7B | 60.6 | - | - | - | - | | [Mulberry](https://github.com/HJYao00/Mulberry) | 7B | 63.1 | - | - | - | - | | [LMM-R1](https://arxiv.org/abs/2503.07536) | 3B | 63.2 | 26.4 | - | - | 41.6 | | [R1-Onevision](https://github.com/Fancy-MLLM/R1-Onevision?tab=readme-ov-file) | 7B | - | 26.2 | - | - | 44.1 | | [MM-Eureka](https://github.com/ModalMinds/MM-EUREKA) | 8B | 67.1 | 22.2 | - | - | 40.4 | | [MM-Eureka](https://github.com/ModalMinds/MM-EUREKA) | 38B | 64.2 | 26.6 | - | - | 48.9 | | **Open-sourced** | | | | | | | | [Ovis2-8b](https://github.com/AIDC-AI/Ovis) | 8B | 71.8 | 25.9 | 39.4 | 42.3 | - | | [MiniCPM-o-2.6](https://github.com/OpenBMB/MiniCPM-o) | 8B | **71.9** | 21.7 | 36.0 | 35.0 | - | | [Qwen2.5-VL](https://github.com/QwenLM/Qwen2.5-VL) (official) | 7B | 68.2 | 25.4 | 47.9 | 41.1 | - | | [Qwen2.5-VL](https://github.com/QwenLM/Qwen2.5-VL) (reproduced) | 7B | 67.5 | 25.6 | 46.8 | 42.5 | 46.9 | | **Ours** | | | | | | | | **MMR1-math-v0** | 7B | 71.0 | **30.2** | **50.8** | **45.1** | **49.8** | ### Quick Start ```python from transformers import Qwen2_5_VLForConditionalGeneration, AutoTokenizer, AutoProcessor from qwen_vl_utils import process_vision_info # default: Load the model on the available device(s) model = Qwen2_5_VLForConditionalGeneration.from_pretrained( "MMR1/MMR1-Math-v0-7B", torch_dtype=torch.bfloat16, attn_implementation="flash_attention_2", device_map="auto", ) # default processer processor = AutoProcessor.from_pretrained("MMR1/MMR1-Math-v0-7B") # Example input messages = [ { "role": "user", "content": [ { "type": "image", "image": "path/to/image.jpeg", }, {"type": "text", "text": "Describe this image."}, ], } ] # Preparation for inference text = processor.apply_chat_template( messages, tokenize=False, add_generation_prompt=True ) image_inputs, video_inputs = process_vision_info(messages) inputs = processor( text=[text], images=image_inputs, videos=video_inputs, padding=True, return_tensors="pt", ) inputs = inputs.to("cuda") # Inference: Generation of the output generated_ids = model.generate(**inputs, max_new_tokens=128) generated_ids_trimmed = [ out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids) ] output_text = processor.batch_decode( generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False ) print(output_text) ```
Batch inference ```python # Sample messages for batch inference messages1 = [ { "role": "user", "content": [ {"type": "image", "image": "file:///path/to/image1.jpg"}, {"type": "image", "image": "file:///path/to/image2.jpg"}, {"type": "text", "text": "What are the common elements in these pictures?"}, ], } ] messages2 = [ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Who are you?"}, ] # Combine messages for batch processing messages = [messages1, messages2] # Preparation for batch inference texts = [ processor.apply_chat_template(msg, tokenize=False, add_generation_prompt=True) for msg in messages ] image_inputs, video_inputs = process_vision_info(messages) inputs = processor( text=texts, images=image_inputs, videos=video_inputs, padding=True, return_tensors="pt", ) inputs = inputs.to("cuda") # Batch Inference generated_ids = model.generate(**inputs, max_new_tokens=128) generated_ids_trimmed = [ out_ids[len(in_ids) :] for in_ids, out_ids in zip(inputs.input_ids, generated_ids) ] output_texts = processor.batch_decode( generated_ids_trimmed, skip_special_tokens=True, clean_up_tokenization_spaces=False ) print(output_texts) ```
## Citation If you find MMR1 useful for your research and applications, please cite using this BibTeX: ```bibtex @misc{MMR1-Math2025, title={MMR1: Advancing the Frontiers of Multimodal Reasoning}, author={Sicong Leng*, Jing Wang*, Jiaxi Li*, Hao Zhang*, Zhiqiang Hu, Boqiang Zhang, Hang Zhang, Yuming Jiang, Xin Li, Fan Wang, Yu Rong, Aixin Sun†, Shijian Lu†}, year={2025}, howpublished={\url{https://github.com/LengSicong/MMR1}}, } ```