Walk the Talk: Bridging the Reasoning-Action Gap for Thinking with Images via Multimodal Agentic Policy Optimization

• URL: http://arxiv.org/abs/2604.06777v1
• Date: Wed, 08 Apr 2026 07:48:07 GMT
• Title: Walk the Talk: Bridging the Reasoning-Action Gap for Thinking with Images via Multimodal Agentic Policy Optimization
• Abstract summary: Multimodal Large Language Models (MLLMs) have incentivized models to think with images'' by actively invoking visual tools during multi-turn reasoning.<n>The common Reinforcement Learning practice of relying on outcome-based rewards ignores the fact that textual plausibility often masks executive failure.<n>We introduce Multimodal Agentic Policy Optimization (MAPO), bridging the gap between textual reasoning and visual actions generated by models within their Multimodal Chain-of-Thought.
• Score: 89.68681087743876
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Recent advancements in Multimodal Large Language Models (MLLMs) have incentivized models to ``think with images'' by actively invoking visual tools during multi-turn reasoning. The common Reinforcement Learning (RL) practice of relying on outcome-based rewards ignores the fact that textual plausibility often masks executive failure, meaning that models may exhibit intuitive textual reasoning while executing imprecise or irrelevant visual actions within their agentic reasoning trajectories. This reasoning-action discrepancy introduces noise that accumulates throughout the multi-turn reasoning process, severely degrading the model's multimodal reasoning capabilities and potentially leading to training collapse. In this paper, we introduce Multimodal Agentic Policy Optimization (MAPO), bridging the gap between textual reasoning and visual actions generated by models within their Multimodal Chain-of-Thought (MCoT). Specifically, MAPO mandates the model to generate explicit textual descriptions for the visual content obtained via tool usage. We then employ a novel advantage estimation that couples the semantic alignment between these descriptions and the actual observations with the task reward. Theoretical findings are provided to justify the rationale behind MAPO, which inherently reduces the variance of gradients, and extensive experiments demonstrate that our method achieves superior performance across multiple visual reasoning benchmarks.

Related papers

• OPLD: On-Policy Latent Distillation for Multimodal Reasoning [27.842942669819948]
  multimodal Chain-of-leaved (CoT) improves visual reasoning by incorporating auxiliary visual evidence into intermediate reasoning.<n>Existing approaches remain constrained by externally defined reasoning traces and visual operations.<n>We propose OPLD, a framework that transfers the reasoning capability induced by privileged multimodal CoT into latent reasoning representations.
  arXiv  Detail & Related papers  (2026-07-30T12:57:45Z)
• MuPHI: Learning Implicit Multimodal Harm Reasoning via Semantically Grounded Reward Optimization [14.713115117792503]
  We introduce Multimodal Pragmatic Harm Interpretation (MuPHI), a dataset containing image-text pairs where harm is encoded in subtle multimodal cues.<n>MuPHIRM is a reasoning-augmented training framework which learns joint semantics by optimizing multi-perspective rewards.<n>Our findings suggest that reasoning-oriented reward optimization offers a promising direction towards building multimodal systems that generalize beyond benchmark-specific shortcuts.
  arXiv  Detail & Related papers  (2026-05-28T13:58:15Z)
• Enhancing Multimodal In-Context Learning via Inductive-Deductive Reasoning [65.15766304205657]
  In-context learning (ICL) allows large models to adapt to tasks using a few examples, yet its extension to vision-language models (VLMs) remains fragile.<n>Our analysis reveals that the fundamental limitation lies in an inductive gap, models often produce correct answers from flawed reasoning.<n>We introduce a framework that restructures multimodal ICL as a principled inductive-deductive process.
  arXiv  Detail & Related papers  (2026-05-04T09:18:19Z)
• Do MLLMs Really See It: Reinforcing Visual Attention in Multimodal LLMs [55.61018839017648]
  Chain-of-thought (CoT) reasoning has substantially improved multimodal large language models (MLLMs) on complex reasoning tasks.<n>Existing approaches largely rely on long textual reasoning trajectories and provide limited mechanisms for learning stable visual attention policies.<n>We propose SAYO, a visual reasoning model trained with a reinforcement learning framework that introduces a region-level visual attention-based reward.
  arXiv  Detail & Related papers  (2026-02-09T03:33:23Z)
• Analyzing Reasoning Consistency in Large Multimodal Models under Cross-Modal Conflicts [74.47786985522762]
  We identify a critical failure mode termed textual inertia, where models tend to blindly adhere to the erroneous text while neglecting conflicting visual evidence.<n>We propose the LogicGraph Perturbation Protocol that structurally injects perturbations into the reasoning chains of diverse LMMs.<n>Results reveal that models successfully self-correct in less than 10% of cases and predominantly succumb to blind textual error propagation.
  arXiv  Detail & Related papers  (2026-01-07T16:39:34Z)
• Reasoning Within the Mind: Dynamic Multimodal Interleaving in Latent Space [46.05748768260013]
  We propose a test-time Dynamic Multimodal Latent Reasoning framework.<n>It employs confidence-guided latent policy gradient optimization to latent think tokens for in-depth reasoning.<n> Experiments across seven multimodal reasoning benchmarks and various model architectures demonstrate that DMLR significantly improves reasoning and perception performance.
  arXiv  Detail & Related papers  (2025-12-14T10:07:45Z)
• Mitigating Visual Forgetting via Take-along Visual Conditioning for Multi-modal Long CoT Reasoning [53.790502697674754]
  We propose Take-along Visual Conditioning (TVC), a strategy that shifts image input to critical reasoning stages.<n>TVC helps the model retain attention to the visual components throughout the reasoning.<n>Our approach achieves state-of-the-art performance on average across five mathematical reasoning benchmarks.
  arXiv  Detail & Related papers  (2025-03-17T16:45:12Z)
• Can MLLMs Reason in Multimodality? EMMA: An Enhanced MultiModal ReAsoning Benchmark [73.27104042215207]
  We introduce EMMA, a benchmark targeting organic multimodal reasoning across mathematics, physics, chemistry, and coding.<n>EMMA tasks demand advanced cross-modal reasoning that cannot be addressed by reasoning independently in each modality.<n>Our evaluation of state-of-the-art MLLMs on EMMA reveals significant limitations in handling complex multimodal and multi-step reasoning tasks.
  arXiv  Detail & Related papers  (2025-01-09T18:55:52Z)
• Cantor: Inspiring Multimodal Chain-of-Thought of MLLM [83.6663322930814]
  We argue that converging visual context acquisition and logical reasoning is pivotal for tackling visual reasoning tasks. We propose an innovative multimodal CoT framework, termed Cantor, characterized by a perception-decision architecture. Our experiments demonstrate the efficacy of the proposed framework, showing significant improvements in multimodal CoT performance.
  arXiv  Detail & Related papers  (2024-04-24T17:59:48Z)

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.