Research map / Vision-language and multimodal / CLIP and zero-shot
What to Trust? A Trust-aware Knowledge-guided Method for Zero-shot Object State Understanding in Videos
Yayun Qi, Xinxiao Wu
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 state temporal dependency extraction to voting (ablation) | trust-aware method · MOST (average over 6 objects) | ATK: 0.42 -> 0.44 | Improved |
| add adaptive knowledge refinement to voting alone (no temporal constraints) (ablation) | trust-aware method · MOST (average over 6 objects) | ATK: 0.42 -> 0.42 | No clear effect |
| add adaptive refinement on top of voting + temporal (ablation) | trust-aware method · MOST (average over 6 objects) | ATK: 0.44 -> 0.49 | Improved |
| add knowledge-guided voting (vs temporal only) (ablation) | trust-aware method · MOST (average over 6 objects) | ATK: 0.34 -> 0.44 | Improved |
| use generative VLM word probabilities instead of CLIP similarity for textual knowledge voting (ablation) | knowledge voting · MOST (average over 6 objects) | Coverage Error: 4.02 -> 2.12 | Improved |
| use soft word-probability scores instead of binary yes/no answers (ablation) | knowledge voting · MOST (average over 6 objects) | ATK: 0.39 -> 0.49 | Improved |
| use word-probability scores instead of VLM-predicted numbers (ablation) | knowledge voting · MOST (average over 6 objects) | ATK: 0.37 -> 0.49 | Improved |
| zero-shot method vs trained LMOSA | object state method · MOST (average over 6 objects) | F1-max: 0.54 -> 0.58 | Improved |
| zero-shot method vs open-vocabulary VIDOSC | object state method · ChangeIT (novel object split) | State Prec@1: 0.29 -> 0.37 | Improved |
| zero-shot method vs trained LMOSA on Egg mAP | object state method · MOST (Egg) | mAP: 0.48 -> 0.49 | No clear effect |
Closest papers by meaning
- Learning Object State Changes in Videos: An Open-World Perspective (CVPR 2024 · similarity 0.88)
- PerceptionLM: Open-Access Data and Models for Detailed Visual Understanding (NeurIPS 2025 · similarity 0.88)
- VideoHallu: Evaluating and Mitigating Multi-modal Hallucinations on Synthetic Video Understanding (NeurIPS 2025 · similarity 0.87)
- Act2See: Emergent Active Visual Perception for Video Reasoning (CVPR 2026 · similarity 0.86)
- MMBench-Video: A Long-Form Multi-Shot Benchmark for Holistic Video Understanding (NeurIPS 2024 · similarity 0.86)
- MVVBench: Benchmarking 4D Reasoning in Vision-Language Models (NeurIPS 2026 · similarity 0.86)
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.
