# Commonsense Video Question Answering through Video-Grounded Entailment Tree Reasoning Source: [Commonsense Video Question Answering through Video-Grounded Entailment Tree Reasoning]() Collection index: [llms.txt]() Data updated: 2026-01-04T04:20:24.020Z Authors: Huabin Liu, Filip Ilievski, Cees G. M. Snoek Published: 2025-01-09T08:44:42.000Z ## Abstract This paper proposes the first video-grounded entailment tree reasoning method for commonsense video question answering (VQA). Despite the remarkable progress of large visual-language models (VLMs), there are growing concerns that they learn spurious correlations between videos and likely answers, reinforced by their black-box nature and remaining benchmarking biases. Our method explicitly grounds VQA tasks to video fragments in four steps: entailment tree construction, video-language entailment verification, tree reasoning, and dynamic tree expansion. A vital benefit of the method is its generalizability to current video and image-based VLMs across reasoning types. To support fair evaluation, we devise a de-biasing procedure based on large-language models that rewrites VQA benchmark answer sets to enforce model reasoning. Systematic experiments on existing and de-biased benchmarks highlight the impact of our method components across benchmarks, VLMs, and reasoning types. ## Recorded benchmark results A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. - [Video Question Answering on NExT-QA de-biased set](): Video Question Answering; NExT-QA de-biased set; Temporal Reasoning Score: 60.8. - [Video Question Answering on IntentQA (de-biased set)](): Video Question Answering; IntentQA (de-biased set); Temporal Score: 61. - [Video Question Answering on IntentQA (original test)](): Video Question Answering; IntentQA original (test); Temporal Accuracy: 66.1. - [Video Question Answering on VideoMME original (test)](): Video Question Answering; VideoMME original (test); Temporal Score: 55.9. - [Video Question Answering on NExT-QA original (test)](): Video Question Answering; NExT-QA original (test); Temporal Score: 64.8. - [Video Question Answering on VideoMME (Aspect Metrics)](): Video Question Answering; VideoMME; VYR: not recorded. ## Original sources and implementation [Original paper]() [PDF]()