Method Overview
Compositional vision–language questions require reasoning over multiple entities and relations. We present Divide and Reason, a framework that jointly decomposes the image and the language into aligned, tractable subproblems. Given a question, our method derives a structured plan and iteratively grounds visual primitives (objects, attributes, relations) while refining the textual spans that guide each step. This joint image–language decomposition improves perceptual grounding, enables robust multi‑hop reasoning, and reduces spurious shortcuts—without changing underlying model architectures. We observe consistent gains in accuracy and faithfulness across diverse benchmarks.
Each example illustrates how the approach decomposes a question into image and language components, grounding entities and relations step‑by‑step.
