Research map / Video / Video-language understanding
Explainable Video Entailment With Grounded Visual Evidence
Junwen Chen, Yu Kong
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add golden negative entities (grounding loss Lg) (ablation) | entity grounding · VIOLIN test | accuracy: 66.53 -> 68.39 | Improved |
| add golden negative entities (grounding loss Lg) (ablation) | entity grounding · VIOLIN test | fake accuracy: 48.01 -> 57.25 | Improved |
| add temporal grounding guidance to entailment judgment (ablation) | entailment judgment · VIOLIN test | accuracy: 67.60 -> 68.39 | Improved |
| add contradiction judgment from entity grounding module (ablation) | entity grounding · VIOLIN test | accuracy: 66.72 -> 68.39 | Improved |
| add explanation regularizer Lr (multi-task) (ablation) | entailment judgment · VIOLIN test | accuracy: 67.32 -> 68.39 | Improved |
| train explanation with only human-written statements instead of also adversarial ones (ablation) | explanation module · VIOLIN test | accuracy: 67.65 -> 68.39 | Improved |
| explain with words instead of constituents (ablation) | explanation module · VIOLIN test | explanation accuracy: 72.42 -> 75.20 | Improved |
| add explanation regularizer Lr (ablation) | entailment judgment · VIOLIN test | real accuracy: 80.63 -> 79.52 | Mixed |
| replace VIOLIN with grounded-evidence model (C3D) | entailment model · VIOLIN test | accuracy: 67.23 -> 68.15 | Improved |
| replace VIOLIN with grounded-evidence model (Resnet) | entailment model · VIOLIN test | accuracy: 67.60 -> 68.39 | Improved |
| replace VIOLIN with grounded-evidence model (Resnet) | entailment model · VIOLIN test | adversarially sampled accuracy: 84.49 -> 84.94 | Improved |
| replace VIOLIN with grounded-evidence model (C3D) | entailment model · VIOLIN test | fake accuracy: 57.73 -> 57.08 | No clear effect |
Closest papers by meaning
- Commonsense Video Question Answering through Video-Grounded Entailment Tree Reasoning (CVPR 2025 · similarity 0.87)
- Defeasible Visual Entailment: Benchmark, Evaluator, and Reward-Driven Optimization (AAAI 2025 · similarity 0.86)
- VITED: Video Temporal Evidence Distillation (CVPR 2025 · similarity 0.83)
- VideoEspresso: A Large-Scale Chain-of-Thought Dataset for Fine-Grained Video Reasoning via Core Frame Selection (CVPR 2025 · similarity 0.83)
- CG-Bench: Clue-grounded Question Answering Benchmark for Long Video Understanding (ICLR 2025 · similarity 0.83)
- VELOCITI: Benchmarking Video-Language Compositional Reasoning with Strict Entailment (CVPR 2025 · similarity 0.83)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
