The field of multimodal large language models (MLLMs) is moving towards improving their reasoning capabilities, particularly in visual text grounding, ordinal understanding, and multimodal explanation. Researchers are exploring novel approaches, such as introducing visual keypoints, Chain-of-Thought (CoT) distillation, and hybrid optimization strategies, to enhance the performance of MLLMs. Noteworthy papers include OrderChain, which presents a prompting paradigm to improve ordinal understanding ability, and Skywork R1V, which introduces a multimodal reasoning model with an efficient multimodal transfer method. Additionally, benchmarks such as MDK12-Bench and V-MAGE are being developed to evaluate the reasoning capabilities of MLLMs in various domains.
Multimodal Reasoning Developments
Sources
Why Reasoning Matters? A Survey of Advancements in Multimodal Reasoning (v1)
Explain with Visual Keypoints Like a Real Mentor! A Benchmark for Multimodal Solution Explanation
OrderChain: A General Prompting Paradigm to Improve Ordinal Understanding Ability of MLLM
Towards Visual Text Grounding of Multimodal Large Language Model
Skywork R1V: Pioneering Multimodal Reasoning with Chain-of-Thought
MDK12-Bench: A Multi-Discipline Benchmark for Evaluating Reasoning in Multimodal Large Language Models
V-MAGE: A Game Evaluation Framework for Assessing Visual-Centric Capabilities in Multimodal Large Language Models
Benchmarking Multimodal CoT Reward Model Stepwise by Visual Program
SCI-Reason: A Dataset with Chain-of-Thought Rationales for Complex Multimodal Reasoning in Academic Areas