MM-HELIX: Boosting Multimodal Long-Chain Reflective Reasoning with Holistic Platform and Adaptive Hybrid Policy Optimization

• URL: http://arxiv.org/abs/2510.08540v2
• Date: Sat, 11 Oct 2025 02:52:01 GMT
• Title: MM-HELIX: Boosting Multimodal Long-Chain Reflective Reasoning with Holistic Platform and Adaptive Hybrid Policy Optimization
• Authors: Xiangyu Zhao, Junming Lin, Tianhao Liang, Yifan Zhou, Wenhao Chai, Yuzhe Gu, Weiyun Wang, Kai Chen, Gen Luo, Wenwei Zhang, Junchi Yan, Hua Yang, Haodong Duan, Xue Yang,
• Abstract summary: Long-chain reflective reasoning is a prerequisite for solving complex real-world problems.<n>We build a benchmark consisting 1,260 samples of 42 challenging synthetic tasks.<n>We generate post-training data and explore learning paradigms for exploiting such data.
• Score: 103.74675519953898
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: While current Multimodal Large Language Models (MLLMs) have demonstrated proficiency in reasoning tasks such as mathematics and logic, their capacity for long-chain reflective reasoning, a prerequisite for solving complex real-world problems, remains largely underexplored. In this work, we first conduct an extensive empirical investigation to evaluate this capability. Leveraging a carefully designed data synthesis engine, we construct MM-HELIX, a multimodal benchmark consisting 1,260 samples of 42 challenging synthetic tasks that require iterative thinking and backtracking. Empirical results on this benchmark reveal that existing MLLMs exhibit significant performance deficits in long-chain reflective reasoning. To address this limitation, we generate post-training data and further explore learning paradigms for exploiting such data. We first develop the Step-Elicited Response Generation pipeline to create MM-HELIX-100K, a large-scale dataset of 100k high-quality, reflective reasoning traces for instruction-tuning stage. Given that standard Reinforcement Learning fails on complex tasks due to sparse reward signals and catastrophic forgetting after Supervised Fine-Tuning, we propose Adaptive Hybrid Policy Optimization (AHPO), a novel training strategy that dynamically unifies offline supervision and online optimization into a single stage. This strategy enables the model to learn from expert data when rewards are sparse and conduct independent exploration once proficient. When applied to the Qwen2.5-VL-7B baseline, our method achieves a +18.6\% accuracy improvement on MM-HELIX benchmark and demonstrates strong generalization with a +5.7\% average performance gain on general mathematic and logic tasks. Our work demonstrate that reflective reasoning in MLLMs can be effectively learned and generalized, paving the way for developing more capable MLLMs.

Related papers

• OPT-Engine: Benchmarking the Limits of LLMs in Optimization Modeling via Complexity Scaling [13.57588221678224]
  Large Language Models (LLMs) have demonstrated impressive progress in optimization modeling.<n>The boundaries of their capabilities in automated formulation and problem solving remain poorly understood.<n>We propose OPT-ENGINE, a benchmark framework designed to evaluate LLMs on optimization modeling with controllable and scalable difficulty levels.
  arXiv  Detail & Related papers  (2026-01-09T09:22:33Z)
• Grounded in Reality: Learning and Deploying Proactive LLM from Offline Logs [72.08224879435762]
  textttLearn-to-Ask is a simulator-free framework for learning and deploying proactive dialogue agents.<n>Our approach culminates in the successful deployment of LLMs into a live, large-scale online AI service.
  arXiv  Detail & Related papers  (2025-10-29T12:08:07Z)
• Large Language Model enabled Mathematical Modeling [2.132096006921049]
  This research investigates the potential of Large Language Models (LLMs) to bridge the formulation gap using natural language understanding and code generation.<n>DeepSeek-R1 is a cost-efficient and high-performing model trained with reinforcement learning.<n>Our methodology includes baseline assessments, the development of a hallucination taxonomy, and the application of mitigation strategies.
  arXiv  Detail & Related papers  (2025-10-22T17:41:42Z)
• MetaLLMix : An XAI Aided LLM-Meta-learning Based Approach for Hyper-parameters Optimization [0.0]
  We propose MetaLLMiX, a framework combining meta-learning, explainable AI, and efficient LLM reasoning.<n>We show that MetaLLMiX achieves competitive or superior performance to traditional HPO methods while drastically reducing computational cost.<n>Our local deployment outperforms prior API-based approaches, achieving optimal results on 5 of 8 tasks, response time reductions of 99.6-99.9%, and the fastest training times on 6 datasets (2.4-15.7x faster)
  arXiv  Detail & Related papers  (2025-09-11T12:06:34Z)
• General-Reasoner: Advancing LLM Reasoning Across All Domains [64.70599911897595]
  Reinforcement learning (RL) has recently demonstrated strong potential in enhancing the reasoning capabilities of large language models (LLMs)<n>We propose General-Reasoner, a novel training paradigm designed to enhance LLM reasoning capabilities across diverse domains.<n>We train a series of models and evaluate them on a wide range of datasets covering wide domains like physics, chemistry, finance, electronics etc.
  arXiv  Detail & Related papers  (2025-05-20T17:41:33Z)
• Observe-R1: Unlocking Reasoning Abilities of MLLMs with Dynamic Progressive Reinforcement Learning [3.364797975300393]
  We present Observe-R1, a novel framework aimed at enhancing the reasoning capabilities of multimodal large language models (MLLMs)<n>We construct the NeuraLadder dataset, which is organized and sampled according to the difficulty and complexity of data samples for RL training.<n>Experiments with the Qwen2.5-VL-3B and Qwen2.5-VL-7B models on 20k samples from the NeuraLadder dataset show that Observe-R1 outperforms a series of larger reasoning models on both reasoning and general benchmarks.
  arXiv  Detail & Related papers  (2025-05-18T14:08:03Z)
• How Difficulty-Aware Staged Reinforcement Learning Enhances LLMs' Reasoning Capabilities: A Preliminary Experimental Study [16.441081996257576]
  This paper presents a rigorous experimental investigation into how difficulty-aware staged reinforcement learning strategies can substantially improve reasoning performance.<n>We show that strategically selecting training data according to well-defined difficulty levels markedly enhances RL optimization.<n>We will open-source our datasets on GitHub and Hugging Face.
  arXiv  Detail & Related papers  (2025-04-01T14:18:38Z)
• Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models [24.45348222168512]
  We propose the reasoning MLLM, Vision-R1, to improve multimodal reasoning capability.<n>Our model achieves an average improvement of $sim$6% across various multimodal math reasoning benchmarks.<n>Vision-R1-7B achieves a 73.5% accuracy on the widely used MathVista benchmark, which is only 0.4% lower than the leading reasoning model, OpenAI O1.
  arXiv  Detail & Related papers  (2025-03-09T20:06:45Z)
• MR-Ben: A Meta-Reasoning Benchmark for Evaluating System-2 Thinking in LLMs [55.20845457594977]
  Large language models (LLMs) have shown increasing capability in problem-solving and decision-making.<n>We present a process-based benchmark MR-Ben that demands a meta-reasoning skill.<n>Our meta-reasoning paradigm is especially suited for system-2 slow thinking.
  arXiv  Detail & Related papers  (2024-06-20T03:50:23Z)
• Uncertainty Aware Learning for Language Model Alignment [97.36361196793929]
  We propose uncertainty-aware learning (UAL) to improve the model alignment of different task scenarios. We implement UAL in a simple fashion -- adaptively setting the label smoothing value of training according to the uncertainty of individual samples. Experiments on widely used benchmarks demonstrate that our UAL significantly and consistently outperforms standard supervised fine-tuning.
  arXiv  Detail & Related papers  (2024-06-07T11:37:45Z)
• Let's reward step by step: Step-Level reward model as the Navigators for Reasoning [64.27898739929734]
  Process-Supervised Reward Model (PRM) furnishes LLMs with step-by-step feedback during the training phase. We propose a greedy search algorithm that employs the step-level feedback from PRM to optimize the reasoning pathways explored by LLMs. To explore the versatility of our approach, we develop a novel method to automatically generate step-level reward dataset for coding tasks and observed similar improved performance in the code generation tasks.
  arXiv  Detail & Related papers  (2023-10-16T05:21:50Z)

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.