Saved in:
Bibliographic Details
Main Authors: Qi, Jianyu, Zou, Ding, Yan, Wenrui, Ma, Rui, Li, Jiaxu, Zheng, Zhijie, Yang, Zhiguo, Zhao, Rongchang
Format: Preprint
Published: 2025
Subjects:
Online Access:https://arxiv.org/abs/2511.06722
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866917070439972864
author Qi, Jianyu
Zou, Ding
Yan, Wenrui
Ma, Rui
Li, Jiaxu
Zheng, Zhijie
Yang, Zhiguo
Zhao, Rongchang
author_facet Qi, Jianyu
Zou, Ding
Yan, Wenrui
Ma, Rui
Li, Jiaxu
Zheng, Zhijie
Yang, Zhiguo
Zhao, Rongchang
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2511_06722
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Revisiting the Data Sampling in Multimodal Post-training from a Difficulty-Distinguish View
Qi, Jianyu
Zou, Ding
Yan, Wenrui
Ma, Rui
Li, Jiaxu
Zheng, Zhijie
Yang, Zhiguo
Zhao, Rongchang
Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
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.
title Revisiting the Data Sampling in Multimodal Post-training from a Difficulty-Distinguish View
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2511.06722