Research map / Video / Video generation
PhyDetEx: Detecting and Explaining the Physical Plausibility of T2V Models
Zeqing Wang, Keze Wang, Lei Zhang
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 |
|---|---|---|---|
| LoRA fine-tune on PID paired videos | Qwen2.5VL-7B · PID Test | F1 score: 75.4 -> 83.5 | Improved |
| LoRA fine-tune on PID paired videos | Qwen2.5VL-7B · Impossible Videos | F1 score: 70.3 -> 87.8 | Improved |
| tell the VLM the video is AI-generated (C1 to C3) (ablation) | InternVL2.5 26B · Impossible Videos | accuracy on implausible videos (%): 65.23 -> 90.47; plausible 78 -> 51.54 | Mixed |
| train on paired plausible/implausible videos instead of non-paired (ablation) | PhyDetEx · Impossible Videos | F1 score: 71.3 -> 87.8 | Improved |
| train on balanced positive and negative samples instead of negatives only (ablation) | PhyDetEx · Impossible Videos | F1 score: 0.0 -> 87.8 | Improved |
| use LoRA instead of full SFT (ablation) | PhyDetEx · PID Test | F1 score: 80.0 -> 83.5 | Improved |
| use SFT-LoRA instead of DPO (ablation) | PhyDetEx · Impossible Videos | F1 score: 70.9 -> 87.8 | Improved |
| physics-aware DPO with PhyDetEx preference data | WanX2.1 1.3B · VideoPhy PC | physical commonsense: 36.16 -> 37.63 | Improved |
| DPO with preference data from vanilla Qwen2.5-VL 7B | WanX2.1 1.3B · VideoPhy PC | physical commonsense: 36.16 -> 35.92 | Hurt |
| newer VLM generation (Qwen3VL vs Qwen2.5VL larger models) | baseline VLMs · PID Test | F1 score: 76.9 -> 77.4 (72B vs 8B) | Improved |
Closest papers by meaning
- WISA: World simulator assistant for physics-aware text-to-video generation (NeurIPS 2025 · similarity 0.89)
- PhyT2V: LLM-Guided Iterative Self-Refinement for Physics-Grounded Text-to-Video Generation (CVPR 2025 · similarity 0.89)
- VideoREPA: Learning Physics for Video Generation through Relational Alignment with Foundation Models (NeurIPS 2025 · similarity 0.88)
- VideoHallu: Evaluating and Mitigating Multi-modal Hallucinations on Synthetic Video Understanding (NeurIPS 2025 · similarity 0.87)
- Towards World Simulator: Crafting Physical Commonsense-Based Benchmark for Video Generation (ICML 2025 · similarity 0.87)
- Physics-Informed Video Generation via Mixture-of-Experts Latent Alignment (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.
