mlsanityData to 7 Oct 2026

MMRPT: MultiModal Reinforcement Pre-Training via Masked Vision-Dependent Reasoning

arXiv 2512.07203 ↗First posted 8 Dec 2025cs.CVcited by 1

On the map: LLM reasoning → Multimodal reasoning

Abstract (arXiv)

Multimodal pre-training remains constrained by the descriptive bias of image-caption pairs, leading models to favor surface linguistic cues over grounded visual understanding. We introduce MMRPT, a masked multimodal reinforcement pre-training framework that strengthens visual reasoning in MLLMs. We are the first to incorporate reinforcement learning directly into the pre-training of large vision-language models, enabling learning signals that reward visual grounding rather than caption imitation. MMRPT constructs masked multimodal data by estimating sentence-level visual dependency via attention over visual tokens and masking highly vision-dependent segments; the model reconstructs these spans through vision-grounded reasoning guided by a semantic-visual reward. Experiments show consistent zero-shot gains across diverse benchmarks and substantially improved robustness under supervised fine-tuning, demonstrating that reinforcement-driven masked reasoning provides a more reliable and generalizable pre-training objective for multimodal models.

Key facts. Cited by 1 paper in the map (data through 7 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 32 of its 33 references that are in the map. First posted on arXiv 8 Dec 2025.

Builds on

32 of 33 references.

Closest ancestors · 4

Same direction, newest first

  1. 2025From Perception to Cognition: A Survey of Vision-Language Interactive Reasoning in Multimodal Large Language Models29 Sep
  2. 2025GLM-4.5V and GLM-4.1V-Thinking: Towards Versatile Multimodal Reasoning with Scalable Reinforcement Learning1 Jul
  3. 2024Thinking Before Looking: Improving Multimodal LLM Reasoning via Mitigating Visual Hallucination15 Nov
  4. 2024Are Large Vision Language Models up to the Challenge of Chart Comprehension and Reasoning? An Extensive Investigation into the Capabilities and Limitations of LVLMs1 Jun

Other references · 28

  1. 2021Learning Transferable Visual Models From Natural Language Supervisioncited by 33,456
  2. 2025Qwen2.5-VL Technical Reportcited by 5,418
  3. 2022LAION-5B: An open large-scale dataset for training next generation image-text modelscited by 4,131
  4. 2023InternVL: Scaling up Vision Foundation Models and Aligning for Generic Visual-Linguistic Taskscited by 2,534
  5. 2024HybridFlow: A Flexible and Efficient RLHF Frameworkcited by 1,980
  6. 2021Align before Fuse: Vision and Language Representation Learning with Momentum Distillationcited by 1,612
  7. 2022CoCa: Contrastive Captioners are Image-Text Foundation Modelscited by 1,265
  8. 2023A Survey on Multimodal Large Language Modelscited by 898
  9. 2024Are We on the Right Way for Evaluating Large Vision-Language Models?cited by 893
  10. 2024Hallucination of Multimodal Large Language Models: A Surveycited by 380
  11. 2025MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipecited by 174
  12. 2024The Revolution of Multimodal Large Language Models: A Surveycited by 134
16 more
  1. 2022Does CLIP Bind Concepts? Probing Compositionality in Large Image Modelscited by 101
  2. 2024Why are Visually-Grounded Language Models Bad at Image Classification?cited by 101
  3. 2024Efficient Multimodal Large Language Models: A Surveycited by 95
  4. 2023LION : Empowering Multimodal Large Language Model with Dual-Level Visual Knowledgecited by 80
  5. 2025Your Large Vision-Language Model Only Needs A Few Attention Heads For Visual Groundingcited by 78
  6. 2024Exploring the Frontier of Vision-Language Models: A Survey of Current Methodologies and Future Directionscited by 66
  7. 2025Vision Language Models are Biasedcited by 63
  8. 2024Multi-Object Hallucination in Vision-Language Modelscited by 60
  9. 2023Revisiting the Role of Language Priors in Vision-Language Modelscited by 50
  10. 2024Looking Beyond Text: Reducing Language bias in Large Vision-Language Models via Multimodal Dual-Attention and Soft-Image Guidancecited by 32
  11. 2025SECOND: Mitigating Perceptual Hallucination in Vision-Language Models via Selective and Contrastive Decodingcited by 19
  12. 2025Modality Bias in LVLMs: Analyzing and Mitigating Object Hallucination via Attention Lenscited by 8
  13. 2025Erase to Improve: Erasable Reinforcement Learning for Search-Augmented LLMscited by 6
  14. 2025Direct Visual Grounding by Directing Attention of Visual Tokenscited by 6
  15. 2025See Different, Think Better: Visual Variations Mitigating Hallucinations in LVLMscited by 3
  16. 2025Learning to Look: Cognitive Attention Alignment with Vision-Language Modelscited by 2

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2026: 1.

1 paper.

  1. 2026Shattering the Autoregressive Curse: Dynamic Epistemic Entropy Orchestrated Erasable Reinforcement Learning for LLMscited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2026See It, Say It, Sorted: An Iterative Training-Free Framework for Visually-Grounded Multimodal Reasoning in LVLMs
  2. 2025Learning Only with Images: Visual Reinforcement Learning with Reasoning, Rendering, and Visual Feedback
  3. 2025Activating Visual Context and Commonsense Reasoning through Masked Prediction in VLMs
  4. 2026Faithful-MR1: Faithful Multimodal Reasoning via Anchoring and Reinforcing Visual Attention
  5. 2026Not All Tokens See Equally: Perception-Grounded Policy Optimization for Large Vision-Language Models
  6. 2026Anchoring What Matters: A Dual-Level Learning Framework for Visually-Grounded Multimodal Reasoning
  7. 2026Visually-Guided Policy Optimization for Multimodal Reasoning
  8. 2025Perception-R1: Advancing Multimodal Reasoning Capabilities of MLLMs via Visual Perception Reward