mlsanityData to 14 Oct 2026

Revisiting the Data Sampling in Multimodal Post-training from a Difficulty-Distinguish View

arXiv 2511.06722 ↗First posted 10 Nov 2025cs.CV cs.AI cs.CLcited by 0

On the map: LLM reasoning → Multimodal reasoning

also in 1 more
Abstract (arXiv)

Recent advances in Multimodal Large Language Models (MLLMs) have spurred significant progress in Chain-of-Thought (CoT) reasoning. Building on the success of Deepseek-R1, researchers extended multimodal reasoning to post-training paradigms based on reinforcement learning (RL), focusing predominantly on mathematical datasets. However, existing post-training paradigms tend to neglect two critical aspects: (1) The lack of quantifiable difficulty metrics capable of strategically screening samples for post-training optimization. (2) Suboptimal post-training paradigms that fail to jointly optimize perception and reasoning capabilities. To address this gap, we propose two novel difficulty-aware sampling strategies: Progressive Image Semantic Masking (PISM) quantifies sample hardness through systematic image degradation, while Cross-Modality Attention Balance (CMAB) assesses cross-modal interaction complexity via attention distribution analysis. Leveraging these metrics, we design a hierarchical training framework that incorporates both GRPO-only and SFT+GRPO hybrid training paradigms, and evaluate them across six benchmark datasets. Experiments demonstrate consistent superiority of GRPO applied to difficulty-stratified samples compared to conventional SFT+GRPO pipelines, indicating that strategic data sampling can obviate the need for supervised fine-tuning while improving model accuracy. Our code will be released at https://github.com/qijianyu277/DifficultySampling.

Key facts. No citing papers in the map yet (data through 14 Oct 2026). Placed in Multimodal reasoning, one of 17 directions in LLM reasoning. Builds on 32 of its 43 references that are in the map. First posted on arXiv 10 Nov 2025.

Builds on

32 of 43 references.

Closest ancestors · 13

Same direction, newest first

  1. 2025Pixel Reasoner: Incentivizing Pixel-Space Reasoning with Curiosity-Driven Reinforcement Learning21 May
  2. 2025Unlocking the Potential of Difficulty Prior in RL-based Multimodal Reasoning19 May
  3. 2025NoisyRollout: Reinforcing Visual Reasoning with Data Augmentation17 Apr
  4. 2025SoTA with Less: MCTS-Guided Sample Selection for Data-Efficient Visual Reasoning Self-Improvement10 Apr
  5. 2025Visual-RFT: Visual Reinforcement Fine-Tuning3 Mar
  6. 2025LlamaV-o1: Rethinking Step-by-step Visual Reasoning in LLMs10 Jan
  7. 2024MAmmoTH-VL: Eliciting Multimodal Reasoning with Instruction Tuning at Scale6 Dec
  8. 2024LLaVA-CoT: Let Vision Language Models Reason Step-by-Step15 Nov
  9. 2024Improve Vision Language Model Chain-of-thought Reasoning21 Oct
  10. 2024Cantor: Inspiring Multimodal Chain-of-Thought of MLLM24 Apr
  11. 2023DDCoT: Duty-Distinct Chain-of-Thought Prompting for Multimodal Reasoning in Language Models25 Oct
  12. 2023MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts3 Oct
1 more
  1. 2023Multimodal Chain-of-Thought Reasoning in Language Models2 Feb

Other references · 19

  1. 2024DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Modelscited by 9,379
  2. 2025DAPO: An Open-Source LLM Reinforcement Learning System at Scalecited by 2,810
  3. 2023MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGIcited by 2,080
  4. 2024LlamaFactory: Unified Efficient Fine-Tuning of 100+ Language Modelscited by 1,464
  5. 2025Understanding R1-Zero-Like Training: A Critical Perspectivecited by 1,345
  6. 2023MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilitiescited by 1,239
  7. 2024Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMscited by 838
  8. 2025Beyond the 80/20 Rule: High-Entropy Minority Tokens Drive Effective Reinforcement Learning for LLM Reasoningcited by 559
  9. 2022GIT: A Generative Image-to-text Transformer for Vision and Languagecited by 537
  10. 2025SimpleRL-Zoo: Investigating and Taming Zero Reinforcement Learning for Open Base Models in the Wildcited by 511
  11. 2025Open-Reasoner-Zero: An Open Source Approach to Scaling Up Reinforcement Learning on the Base Modelcited by 461
  12. 2024SWIFT:A Scalable lightWeight Infrastructure for Fine-Tuningcited by 355
7 more
  1. 2025VAPO: Efficient and Reliable Reinforcement Learning for Advanced Reasoning Taskscited by 285
  2. 2025A Minimalist Approach to LLM Reasoning: from Rejection Sampling to Reinforcecited by 145
  3. 2024Superfiltering: Weak-to-Strong Data Filtering for Fast Instruction-Tuningcited by 138
  4. 2025GRPO-LEAD: A Difficulty-Aware Reinforcement Learning Approach for Concise Mathematical Reasoning in Language Modelscited by 82
  5. 2025R1-Reward: Training Multimodal Reward Model Through Stable Reinforcement Learningcited by 65
  6. 2025DUMP: Automated Distribution-Level Curriculum Learning for RL-based LLM Post-trainingcited by 46
  7. 2025One RL to See Them All: Visual Triple Unified Reinforcement Learningcited by 30

Built on by

Not cited yet.

Similar, not cited

Close in wording, no citation either way

  1. 2025Observe-R1: Unlocking Reasoning Abilities of MLLMs with Dynamic Progressive Reinforcement Learning
  2. 2025Advancing Multimodal Reasoning: From Optimized Cold Start to Staged Reinforcement Learning
  3. 2025MMRPT: MultiModal Reinforcement Pre-Training via Masked Vision-Dependent Reasoning
  4. 2025Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models
  5. 2026Reinforcing Multimodal Reasoning via Token-Level Perception-Grounded Advantage Estimation
  6. 2026See It, Say It, Sorted: An Iterative Training-Free Framework for Visually-Grounded Multimodal Reasoning in LVLMs
  7. 2026Reinforcing Multimodal Reasoning Against Visual Degradation
  8. 2025The Synergy Dilemma of Long-CoT SFT and RL: Investigating Post-Training Techniques for Reasoning VLMs