Research map / Video / Video-language understanding
Dr.V : A Hierarchical Perception-Temporal-Cognition Framework to Diagnose Video Hallucination by Fine-Grained Spatial-Temporal Grounding
Meng Luo, Shengqiong Wu, Liqiang Jing, Tianjie Ju, Zheng Li, Jinxiang Lai, Tianlong Wu, Xinya Du, Jian Li, Siyuan Yan, Jiebo Luo, William Yang Wang and 3 more
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 Dr.V-Agent tool-based diagnosis | VideoChat2 · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): 38.01 -> 53.43 (+15.42) | Improved |
| add Dr.V-Agent tool-based diagnosis | LLaVA-NeXT · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): 56.80 -> 74.21 (+17.41) | Improved |
| add Dr.V-Agent tool-based diagnosis | Qwen2-VL · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): 72.67 -> 82.64 (+9.97) | Improved |
| add Dr.V-Agent tool-based diagnosis | GPT-4o · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): 77.29 -> 88.36 (+11.07) | Improved |
| add Self-PEP self-correction | LLaVA-NeXT · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): 56.80 -> 52.45 (-4.35) | Hurt |
| add Self-PEP self-correction | VideoChat2 · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): 38.01 -> 44.82 (+6.81) | Improved |
| add Self-PEP self-correction | Qwen2-VL · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): 72.67 -> 75.67 (+3.00) | Improved |
| add Self-PEP self-correction | GPT-4o · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): 77.29 -> 82.33 (+5.04) | Improved |
| add Self-PEP self-correction | PLLaVA · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): -5.30% | Hurt |
| replace Self-PEP with Dr.V-Agent | VideoChat2 · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): +18.60% | Improved |
| replace Self-PEP with Dr.V-Agent | GPT-4o · Dr.V-Bench (Self-PEP comparison) | average accuracy (%): 82.33 -> 88.36 | Improved |
| move from perceptive to temporal question types | Qwen2-VL · Dr.V-Bench (multiple-choice QA) | accuracy (%): 78.75 -> 65.61 | Hurt |
Closest papers by meaning
- VidHalluc: Evaluating Temporal Hallucinations in Multimodal Large Language Models for Video Understanding (CVPR 2025 · similarity 0.91)
- MHBench: Demystifying Motion Hallucination in VideoLLMs (AAAI 2025 · similarity 0.90)
- Seeing Is Believing: Rich-Context Hallucination Detection for MLLMs via Backward Visual Grounding (AAAI 2026 · similarity 0.90)
- Mitigating Hallucination in VideoLLMs via Temporal-Aware Activation Engineering (NeurIPS 2025 · similarity 0.90)
- Learning to Decode Against Compositional Hallucination in Video Multimodal Large Language Models (ICML 2026 · similarity 0.89)
- Causally-Grounded Dual-Path Attention Intervention for Object Hallucination Mitigation in LVLMs (AAAI 2026 · similarity 0.89)
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.
