Decompose, Look, and Reason: Reinforced Latent Reasoning for VLMs

TL;DR

The paper introduces DLR, a novel latent reasoning framework for vision-language models that decomposes queries, extracts visual latents, and improves reasoning accuracy and interpretability.

Contribution

DLR is a new framework that enhances visual reasoning by dynamically decomposing queries and grounding visual latents, with a novel training pipeline and latent policy.

Findings

DLR outperforms existing methods on vision-centric benchmarks.

DLR provides superior stepwise interpretability.

DLR demonstrates effective exploration in latent space.

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 \emph{"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.