Answer Self-Consistency with Margin-Triggered Question Re-Arbitration for the CVPR 2026 VidLLMs Challenge
2026-06-03 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors developed a method called ASC-MQRA to help computers better understand relationships in videos by asking questions multiple times and checking for consistent answers. Their main idea, ASC, runs several question-answer attempts and picks answers that appear most often, improving accuracy compared to just one try. They also tested an extra step, MQRA, to double-check uncertain cases by focusing on the most likely answers, which helped during development but didn’t improve their final test results. Their final model, using only ASC, performed well on a video relational reasoning challenge. The report explains their techniques, experiments, and provides the code for others to use.
Visual Relational ReasoningVideo Question AnsweringMultimodal ReasoningSelf-ConsistencyTest-Time ReasoningVote MarginUncertainty EstimationRe-ArbitrationPrompting StrategyEnsemble Methods
Authors
Tomoya Miyazawa, Hiroyasu Okuno
Abstract
In this report, we present our solution for Track 2 of the CVPR 2026 VidLLMs Challenge. This track evaluates visual relational reasoning in videos, where models must infer relations that are not always explicitly visible. We propose Answer Self-Consistency with Margin-Triggered Question Re-Arbitration (ASC-MQRA), a training-free test-time reasoning framework built on a multimodal reasoning model. The core ASC component performs multiple stochastic video question-answering runs and aggregates their answer choices through answer-level self-consistency. This substantially improves over single-pass inference and forms our final test submission. We further study MQRA, a conditional re-arbitration module for low-margin examples where the first-stage vote distribution indicates uncertainty. Our vote-margin analysis shows that low-margin examples often retain the ground-truth answer among the top candidates, motivating MQRA to narrow the candidate set and re-watch the video only over the retained candidates. On validation, MQRA further improves over ASC, indicating that low-margin vote distributions can provide a useful uncertainty signal. On test, however, MQRA slightly degrades performance relative to ASC, suggesting that re-arbitration is sensitive to the size and category distribution of the triggered subset. Our final test submission therefore uses ASC without re-arbitration, achieving 72.73 average accuracy and 78.34 category-wise macro average accuracy on validation, and 81.16 average accuracy and 80.91 category-wise macro average accuracy on test. This report details our prompting strategy, implementation setup, ablation studies, and diagnostic analyses. The code is available at https://github.com/data-analytics-labo/ASC-MQRA