Training Multimodal Large Reasoning Models Needs Better Thoughts: A Three-Stage Framework for Long Chain-of-Thought Synthesis and Selection

• URL: http://arxiv.org/abs/2512.18956v1
• Date: Mon, 22 Dec 2025 02:07:20 GMT
• Title: Training Multimodal Large Reasoning Models Needs Better Thoughts: A Three-Stage Framework for Long Chain-of-Thought Synthesis and Selection
• Authors: Yizhi Wang, Linan Yue, Min-Ling Zhang,
• Abstract summary: Large Reasoning Models (LRMs) have demonstrated remarkable performance on complex reasoning tasks through long Chain-of-Thought (CoT) reasoning.<n>Existing multimodal datasets and CoT methods still suffer from limited reasoning depth, modality conversion errors, and rigid generation pipelines.<n>We propose SynSelect, a novel three-stage Synthesis-Selection framework for generating high-quality long CoT data tailored to multimodal reasoning tasks.
• Score: 64.34737012956182
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large Reasoning Models (LRMs) have demonstrated remarkable performance on complex reasoning tasks through long Chain-of-Thought (CoT) reasoning. Extending these successes to multimodal reasoning remains challenging due to the increased complexity of integrating diverse input modalities and the scarcity of high-quality long CoT training data. Existing multimodal datasets and CoT synthesis methods still suffer from limited reasoning depth, modality conversion errors, and rigid generation pipelines, hindering model performance and stability. To this end, in this paper, we propose SynSelect, a novel three-stage Synthesis-Selection framework for generating high-quality long CoT data tailored to multimodal reasoning tasks. Specifically, SynSelect first leverages multiple heterogeneous multimodal LRMs to produce diverse candidate CoTs, and then applies both instance and batch level selection to filter high-quality CoTs that can effectively enhance the model's reasoning capabilities. Extensive experiments on multiple multimodal benchmarks demonstrate that models supervised fine-tuned on SynSelect-generated data significantly outperform baselines and achieve further improvements after reinforcement learning post-training. Our results validate SynSelect as an effective approach for advancing multimodal LRMs reasoning capabilities.

Related papers

• Embed-RL: Reinforcement Learning for Reasoning-Driven Multimodal Embeddings [44.77164359074224]
  Multimodal Large Language Models (MLLMs) have become pivotal for advancing Universal Multimodal Embeddings (UME)<n>Recent studies demonstrate that incorporating generative Chain-of-Thought (CoT) reasoning can substantially enhance task-specific representations.<n>We propose a reasoning-driven UME framework that integrates Embedder-Guided Reinforcement Learning (EG-RL) to optimize the Reasoner to produce evidential Traceability CoT.
  arXiv  Detail & Related papers  (2026-02-14T15:35:03Z)
• UniT: Unified Multimodal Chain-of-Thought Test-time Scaling [85.590774707406]
  Unified models can handle both multimodal understanding and generation within a single architecture, yet they typically operate in a single pass without iteratively refining their outputs.<n>We introduce UniT, a framework for multimodal test-time scaling that enables a single unified model to reason, verify, and refine across multiple rounds.
  arXiv  Detail & Related papers  (2026-02-12T18:59:49Z)
• Dual-Phase LLM Reasoning: Self-Evolved Mathematical Frameworks [48.105258051884384]
  This paper proposes a new two-stage training framework that enhances models' self-correction capabilities.<n>During the first stage, a multi-turn dialogue strategy guides the model to generate long chain-of-thought (CoT) data.<n>The second stage employs a difficulty-aware rejection sampling mechanism to dynamically optimize data distribution.
  arXiv  Detail & Related papers  (2026-01-09T08:19:11Z)
• Resource-Limited Joint Multimodal Sentiment Reasoning and Classification via Chain-of-Thought Enhancement and Distillation [22.722731231389393]
  Current approaches primarily leverage the knowledge and reasoning capabilities of parameter-heavy (Multimodal) Large Language Models (LLMs)<n>We propose a Multimodal Chain-of-Student Reasoning Distillation model, MulCoT-RD, to address deployment constraints in resource-limited environments.<n>Experiments on four datasets demonstrate that MulCoT-RD with only 3B parameters achieves strong performance on JMSRC, while exhibiting robust generalization and enhanced interpretability.
  arXiv  Detail & Related papers  (2025-08-07T10:23:14Z)
• FindRec: Stein-Guided Entropic Flow for Multi-Modal Sequential Recommendation [57.577843653775]
  We propose textbfFindRec (textbfFlexible unified textbfinformation textbfdisentanglement for multi-modal sequential textbfRecommendation)<n>A Stein kernel-based Integrated Information Coordination Module (IICM) theoretically guarantees distribution consistency between multimodal features and ID streams.<n>A cross-modal expert routing mechanism that adaptively filters and combines multimodal features based on their contextual relevance.
  arXiv  Detail & Related papers  (2025-07-07T04:09:45Z)
• Unified Multimodal Chain-of-Thought Reward Model through Reinforcement Fine-Tuning [31.727984223052648]
  This paper proposes UnifiedReward-Think, the first unified multimodal CoT-based reward model.<n>We first use a small amount of image generation preference data to distill the reasoning process of GPT-4o.<n>We then prepare large-scale unified multimodal preference data to elicit the model's reasoning process across various vision tasks.
  arXiv  Detail & Related papers  (2025-05-06T08:46:41Z)
• FastMCTS: A Simple Sampling Strategy for Data Synthesis [67.60823802317141]
  We introduce FastMCTS, an innovative data synthesis strategy inspired by Monte Carlo Tree Search.<n>FastMCTS provides a more efficient sampling method for multi-step reasoning data, offering step-level evaluation signals.<n>Experiments on both English and Chinese reasoning datasets demonstrate that FastMCTS generates over 30% more correct reasoning paths.
  arXiv  Detail & Related papers  (2025-02-17T06:27:57Z)
• Progressive Multimodal Reasoning via Active Retrieval [64.74746997923967]
  Multi-step multimodal reasoning tasks pose significant challenges for large language models (MLLMs)<n>We propose AR-MCTS, a universal framework designed to progressively improve the reasoning capabilities of MLLMs.<n>We show that AR-MCTS can optimize sampling diversity and accuracy, yielding reliable multimodal reasoning.
  arXiv  Detail & Related papers  (2024-12-19T13:25:39Z)
• LATTE: Learning to Think with Vision Specialists [110.43838069105998]
  We propose LATTE, a family of vision-language models that offload perception to state-of-the-art vision models.<n>By offloading perception to state-of-the-art vision models, our approach enables vision-language models to focus solely on reasoning over high-quality perceptual information.
  arXiv  Detail & Related papers  (2024-12-07T00:42:04Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.