Seeing Parts, Reasoning about Worlds: Visual Inference under Partial Observation

2026-09-26

AI Summary

This study addresses the challenge of reasoning about complete world states from limited visual evidence under partial observability. To this end, it introduces the WorldScope dataset and the WorldFlow model, which leverage possible-world semantics to learn visual representations. Specifically, the method employs counter-world sets to provide witnesses for physical ambiguity and constructs evidence lattices over image subsets to integrate cross-view information, thereby enabling view-induced world exclusion reasoning. By combining complementary supervision, evidence grounding, and cross-view entity composition mechanisms, the proposed approach achieves an accuracy of 64.34% on the WorldScope-Bench benchmark, outperforming baselines by 24.88 percentage points. Notably, it maintains a robust accuracy of 50.43% even in structurally disjoint scenarios.

Abstract

World modeling under partial observation requires reasoning about the complete worlds that remain compatible with limited visual evidence. Occluded objects and unseen regions can leave several world states possible; additional views can exclude alternatives and strengthen the conclusions supported by the observations. We introduce WorldScope to study this process through possible-world semantics, evidence-grounded data, and learned visual representations. WorldScope-1.2M provides 1.2 million English question-answer pairs spanning eight world properties, ten task interfaces, and three observation protocols. Its answers encode confirmed facts, supported bounds, and unresolved possibilities. Complementary supervision comprises 4,800 certified counterworld groups with equivalent base observations and different hidden object configurations and query answers. These groups provide physical witnesses of ambiguity and training-only labels for world compatibility and view-induced exclusions. We propose WorldFlow, which composes cross-view entity evidence and support-surface coverage into an image-subset evidence lattice. Counterworld compatibility and transition objectives train subset representations to reflect how new observations constrain possible worlds. A shared answer generator uses these representations to predict the strongest supported conclusion. WorldScope-Bench evaluates claim judgments and evidence-dependent conclusions as views are selected, combined, removed, or ordered. On its 5,000-question test set, WorldFlow reaches 64.34% exact accuracy, improving over the same backbone trained on QA alone by 24.88 percentage points. It retains 50.43% accuracy on the 3,000 questions from structure-disjoint scenes.

Problem

• partial observation
• world modeling
• visual reasoning
• occlusion
• possible-world semantics

Innovation

• Partial Observation
• Possible-World Semantics
• Counterworld Supervision
• Evidence Lattice
• Visual Reasoning