Multimodal Reasoning Comparison Hub

Reference Surfaces

Top Papers

• Think While Watching: Online Streaming Segment-Level Memory for Multi-Turn Video Reasoning in Multimodal Large Language Models(9.0)

  A memory-anchored framework for real-time multi-turn video reasoning in multimodal large language models.

• M$^3$-ACE: Rectifying Visual Perception in Multimodal Math Reasoning via Multi-Agentic Context Engineering(8.0)

  M3-ACE is a multi-agent system that improves visual math reasoning by rectifying visual perception, achieving state-of-the-art results and offering a clear path to commercial applications.

• BRIDGE: Benchmark for multi-hop Reasoning In long multimodal Documents with Grounded Evidence(7.0)

  BRIDGE is a benchmark dataset for evaluating multi-hop reasoning in long multimodal documents, enabling targeted diagnosis of reasoning failures in LLMs and RAG systems.

• ClueTracer: Question-to-Vision Clue Tracing for Training-Free Hallucination Suppression in Multimodal Reasoning(7.0)

  ClueTracer enhances multimodal reasoning models by suppressing hallucinations without additional training.

• Fuel Gauge: Estimating Chain-of-Thought Length Ahead of Time in Large Multimodal Models(7.0)

  Fuel Gauge predicts Chain-of-Thought length in multimodal models to optimize resource use and accuracy.

• Concise Geometric Description as a Bridge: Unleashing the Potential of LLM for Plane Geometry Problem Solving(6.0)

  Enhance plane geometry problem solving by converting visual diagrams into concise textual descriptions for LLMs.

• GeoSense: Internalizing Geometric Necessity Perception for Multimodal Reasoning(6.0)

  GeoSense enhances multimodal reasoning by autonomously engaging geometric features based on perceptual necessity.

• Evolving from Tool User to Creator via Training-Free Experience Reuse in Multimodal Reasoning(5.0)

  Create adaptive, self-updating multimodal reasoning tools without training through a unique experience reuse framework.

• Omni-R1: Towards the Unified Generative Paradigm for Multimodal Reasoning(2.0)

  Unified generative paradigm for diverse multimodal reasoning tasks.

• Deconstructing Multimodal Mathematical Reasoning: Towards a Unified Perception-Alignment-Reasoning Paradigm(2.0)

  Develops a framework for integrating perception, alignment, and reasoning to improve multimodal mathematical reasoning models.