Unleashing Perception-Time Scaling to Multimodal Reasoning Models

2025/10/10 by Yifan Li, Li, Yifan, Zhenghao Chen +17 · 1 citation
Computer Science · #Computation and Language (cs.CL) #Computer Vision and Pattern Recognition (cs.CV) #Domain Adaptation and Few-Shot Learning #Explainable Artificial Intelligence (XAI) #FOS: Computer and information sciences #Multimodal Machine Learning Applications

paper · pdf · doi:10.48550/arxiv.2510.08964

openalex publication_date 2025/10/10 · openalex created_date 2025/10/14 · openalex updated_date 2026/07/28

Abstract

Recent advances in inference-time scaling, particularly those leveraging reinforcement learning with verifiable rewards, have substantially enhanced the reasoning capabilities of Large Vision-Language Models (LVLMs). Inspired by this success, similar strategies have been applied to multimodal reasoning, yet their impact on visual perception remains unclear. To investigate this gap, we introduce DisTANCE, a perception-centric benchmark for visual estimation tasks. Evaluation results show that LVLMs exhibit limited estimation precision, and inference-time scaling offers only marginal gains. We attribute this to the fast perception paradigm of current LVLMs, where visual understanding is treated as a one-shot output without modeling the underlying perceptual process. To address this, we propose Perception-Time Scaling (PTS), a novel paradigm that encourages token-rich perception and decomposes complex perception problems into intermediate tractable sub-problems, thereby enabling perception to align with and benefit from inference-time scaling. Combined with reinforcement learning techniques, PTS significantly improves perception accuracy, raising high-precision performance on DisTANCE from 8.0% to 64.7%, and generalizes well to out-of-domain tasks. Surprisingly, even though PTS data are purely synthetic, combining them with math reasoning data yields consistent gains in both reasoning and real-world perception benchmarks. Further analysis reveals that PTS introduces more perception-related tokens and increases the model's attention to image tokens. Our code and data will be publicly released.

Citations

• Are Reasoning Models More Prone to Hallucination?
• More Thinking, Less Seeing? Assessing Amplified Hallucination in Multimodal Reasoning Models
• NoisyRollout: Reinforcing Visual Reasoning with Data Augmentation
• InternVL3: Exploring Advanced Training and Test-Time Recipes for Open-Source Multimodal Models
• SFT or RL? An Early Investigation into Training R1-Like Reasoning Large Vision-Language Models
• LEGO-Puzzles: How Good Are MLLMs at Multi-Step Spatial Reasoning?
• Vision-R1: Evolving Human-Free Alignment in Large Vision-Language Models via Vision-Guided Reinforcement Learning
• R1-Onevision: Advancing Generalized Multimodal Reasoning through Cross-Modal Formalization
• Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models
• Visual-RFT: Visual Reinforcement Fine-Tuning
• Qwen2.5-VL Technical Report
• DeepSeek-R1 incentivizes reasoning in LLMs through reinforcement learning
• Valley2: Exploring Multimodal Models with Scalable Vision-Language Design
• Virgo: A Preliminary Exploration on Reproducing o1-like MLLM
• Mulberry: Empowering MLLM with o1-like Reasoning and Reflection via Collective Monte Carlo Tree Search
• Thinking in Space: How Multimodal Large Language Models See, Remember, and Recall Spaces
• Euclid: Supercharging Multimodal LLMs with Synthetic High-Fidelity Visual Descriptions
• LLaVA-CoT: Let Vision Language Models Reason Step-by-Step
• Cambrian-1: A Fully Open, Vision-Centric Exploration of Multimodal LLMs
• Evaluating and Analyzing Relationship Hallucinations in Large Vision-Language Models
• Visual Sketchpad: Sketching as a Visual Chain of Thought for Multimodal Language Models
• SpatialRGPT: Grounded Spatial Reasoning in Vision Language Models
• BLINK: Multimodal Large Language Models Can See but Not Perceive
• DetToolChain: A New Prompting Paradigm to Unleash Detection Ability of MLLM
• Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset
• BioXP-0.5B: Explainable Medical-AI via RL-GRPO
• SpatialVLM: Endowing Vision-Language Models with Spatial Reasoning Capabilities
• Eyes Wide Shut? Exploring the Visual Shortcomings of Multimodal LLMs
• MMMU: A Massive Multi-discipline Multimodal Understanding and Reasoning Benchmark for Expert AGI
• HallusionBench: An Advanced Diagnostic Suite for Entangled Language Hallucination and Visual Illusion in Large Vision-Language Models
• MM-Vet: Evaluating Large Multimodal Models for Integrated Capabilities
• MMBench: Is Your Multi-modal Model an All-around Player?
• Evaluating Object Hallucination in Large Vision-Language Models
• A Survey of Large Language Models
• Inter-GPS: Interpretable Geometry Problem Solving with Formal Language and Symbolic Reasoning

Cited by

• L2V-CoT: Cross-Modal Transfer of Chain-of-Thought Reasoning via Latent Intervention

Related

• Limits and Gains of Test-Time Scaling in Vision-Language Reasoning
• Perception Before Reasoning: Two-Stage Reinforcement Learning for Visual Reasoning in Vision-Language Models
• BLINK-Twice: You see, but do you observe? A Reasoning Benchmark on Visual Perception
• Perception-R1: Advancing Multimodal Reasoning Capabilities of MLLMs via Visual Perception Reward
• More Thought, Less Accuracy? On the Dual Nature of Reasoning in Vision-Language Models