Disagreement-Based Cross-Model Routing for Implicit Video Question Answering
Quick Answer
The study introduces disagreement-based cross-model routing for video question answering, enhancing accuracy by 1.43% on the ImplicitQA benchmark using Gemini 3.1 Pro Preview and Claude Opus 4.8 models.
Quick Take
This method effectively identifies and routes challenging questions, achieving significant gains in categories reliant on cross-shot references.
Key Points
- Disagreement-based routing improves video QA accuracy on ImplicitQA benchmark by +1.43%.
- Utilizes Gemini 3.1 Pro Preview and Claude Opus 4.8 models for enhanced performance.
- Significant gains observed in Motion & Trajectory (+5.49%) and Inferred Counting (+3.45%).
- Method requires no labels or training, operating purely at inference time.
- Achieved 82.03 AvgAcc on the CVPR 2026 ImplicitQA challenge test set.
Paper Resources
Source Excerpt
arXiv:2606. 14723v1 Announce Type: new Abstract: We study multiple-choice video question answering on the ImplicitQA benchmark, where the correct answer is never explicitly shown but must be inferred from off-screen events, line-of-sight cues, causal structure, and cross-shot spatial layout.
On this benchmark a single frontier video already operates near its accuracy ceiling, and we observe that conventional self-consistency strategies -- majority voting across repeated samples of the same model -- can hurt rather than help, because the model's errors on hard questions are correlated. …
Want this in your inbox every morning?
Daily brief at your local 8am — bilingual EN/中文, free.
More from arXiv cs.CV
See more →ProMoE-FL: Prototype-conditioned Mixture of Experts for Multimodal Federated Learning with Missing Modalities
ProMoE-FL introduces a Prototype-conditioned Mixture-of-Experts framework for multimodal federated learning, effectively addressing missing modalities. It outperforms existing methods on four chest X-ray datasets, demonstrating superior feature synthesis capabilities in both homogeneous and heterogeneous settings.