Decompose, Look, and Reason: Reinforced Latent Reasoning for VLMs
cs.CL
Submitted: 2026-04-08
Updated: 2026-08-30
Comments: Accepted to the EMNLP 2026 Main Conference
License: http://creativecommons.org/licenses/by/4.0/
The gist: Vision-Language Models often struggle with complex visual reasoning due to the visual information loss in textual CoT.
Terminology
Abstract
Vision-Language Models often struggle with complex visual reasoning due to the visual information loss in textual CoT. Existing methods either add the cost of tool calls or rely on localized patch-based embeddings that are insufficient to extract semantics in multi-step reasoning. We propose "Decompose, Look, and Reason" (DLR), a reinforced latent reasoning framework that dynamically decomposes queries into textual premises, extracts premise-conditioned continuous visual latents, and deduces answers through grounded rationales. We introduce a three-stage training pipeline and propose a novel Spherical Gaussian Latent Policy, to enable effective exploration in the latent space. Extensive experiments on vision-centric benchmarks show that DLR consistently outperforms strong baselines, including text-only, interleaved multimodal CoT, and latent reasoning methods, while providing superior stepwise interpretability.
Sources
- Latent Visual Reasoning
- MathVista: Evaluating Mathematical Reasoning of Foundation Models in Visual Contexts
- Qwen3-VL Technical Report
- ReFocus: Visual Editing as a Chain of Thought for Structured Image Understanding
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- OpenThinkIMG: Learning to Think with Images via Visual Tool Reinforcement Learning
- Thinking with Images for Multimodal Reasoning: Foundations, Methods, and Future Frontiers
- Vision-R1: Incentivizing Reasoning Capability in Multimodal Large Language Models
- VL-Rethinker: Incentivizing Self-Reflection of Vision-Language Models with Reinforcement Learning
- Pixel Reasoner: Incentivizing Pixel-Space Reasoning with Curiosity-Driven Reinforcement Learning
- Monet: Reasoning in Latent Visual Space Beyond Images and Language
- Multimodal Chain-of-Thought Reasoning: A Comprehensive Survey
- Exploring the Reasoning Abilities of Multimodal Large Language Models (MLLMs): A Comprehensive Survey on Emerging Trends in Multimodal Reasoning
- Machine Mental Imagery: Empower Multimodal Reasoning with Latent Visual Tokens
- Multimodal Chain-of-Thought Reasoning in Language Models
Related papers
- Exploring Solution Divergence and Its Effect on Large Language Model Problem Solving
- Ishigaki-IDS-Bench: A Benchmark for Generating Information Delivery Specification from BIM Information Requirements
- Subliminal Steering: Stronger Encoding of Hidden Signals
- MedStruct-S: A Benchmark for Key Discovery, Key-Conditioned QA and Semi-Structured Extraction from OCR Clinical Reports
- The End of Transformers? On Challenging Attention and the Rise of Sub-Quadratic Architectures
- Untangling the Mechanisms of Misleading Context in Medical Question Answering