Symmetry-Aware Likelihood-Orbit Aggregation for Selective Left-Right Claim Verification

2026-09-15

AI Summary

针对细粒度左右声明验证问题，提出Relation-Orbit方法，通过反射、逆关系和实体交换生成八个标准化似然值，并使用闭式对比来提高验证信号的选择性。

Abstract

Frozen vision-language models (VLMs) remain unreliable on fine-grained left-right claims, and raw claim likelihoods need not reliably rank verification errors. After a horizontal-reflection intervention is fixed, how should its induced likelihood measurements be combined into a selective verification signal? We introduce Relation-Orbit, a closed-form contrast with no learned fusion parameters that assigns eight normalized likelihoods to query-supporting and counterfactual roles determined by reflection, inverse relation, and entity exchange. A claim is asserted only when the signed contrast exceeds a threshold selected on held-out data using pointwise Clopper-Pearson upper confidence bounds. On VSR and GQA across four frozen VLMs, Relation-Orbit yields higher mean test coverage at a 10% selective-risk calibration target than an all-eight Orbit-Max baseline in all eight dataset-backbone settings; gains over a nearly abstain-all one-sided intervention score are reported separately. A separate LLaVA-1.5/COCO evaluation, reduced-orbit controls, and a two-sided partition diagnostic further characterize the structural advantage.

Problem

• Frozen vision-language models
• Left-right claims
• Likelihood measurements
• Selective verification signal
• Horizontal-reflection intervention

Innovation

• Relation-Orbit
• Likelihood-Contrast
• Selective Verification
• Frozen VLMs
• Reflection Intervention