AI Paper Atlas

Research map / Vision-language and multimodal / Multimodal LLMs

From Pixels to Logic: A Perception-Reasoning Decomposition Framework for Open-World Referring Expression Comprehension

Lihong Huang, Sheng-hua Zhong, Zhi Zhang, Yan Liu

AAAI 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Chain-of-thought multimodal reasoning · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a perception-reasoning decomposition framework for open-world referring expression comprehension, where expressions may refer to zero, one, or multiple objects.

Problem

Referring expression comprehension usually assumes each expression refers to exactly one known object; real-world expressions may refer to multiple objects, none, or novel categories with complex semantics.

Approach

Decomposes open-world REC into a perception stage that proposes candidate objects and a reasoning stage that resolves the expression's logic over them, allowing zero, one, or many referents.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
enable explicit reasoning (thinking mode) (ablation)Qwen3-32B · RefCOCO testAaccuracy: 69.8 -> 78.9Improved
enable explicit reasoning (thinking mode) (ablation)Qwen3-235B-A22B · RefCOCO testAaccuracy: 71.4 -> 78.4Improved
add image caption (ablation)PRDF · gRefCOCO testBprecision: 42.2 -> 52.1Improved
add containment relations (ablation)PRDF · gRefCOCO testBprecision: 45.0 -> 52.1Improved
add depth information (ablation)PRDF · gRefCOCO testBprecision: 52.1 -> 52.1No clear effect
add all three scene representations (ablation)PRDF · gRefCOCO testBprecision: 37.1 -> 52.1Improved
add prompt note about small-region hallucination (ablation)PRDF · RefCOCO testAaccuracy: 80.7 -> 81.3Improved
use more generated masks (ablation)PRDF · gRefCOCOF1: Pearson r between -0.10 and -0.05Hurt
use reasoning LM Deepseek-R1 vs non-reasoning Deepseek-V3 (ablation)PRDF · RefCOCO testAaccuracy: 77.6 -> 80.7Improved
compare zero-shot PRDF with supervised SimVG on no-target accuracySimVG · gRefCOCO valno-target accuracy: 54.7 -> 85.2Improved
compare zero-shot PRDF with supervised SimVG on precisionSimVG · gRefCOCOprecision: 62.1 -> 65.4 (val); 54.8 -> 52.1 (testB)Mixed

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy