Research map / Video / Video-language understanding
VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning
Hao Tan, jun lan, Senyuan Shi, Zichang Tan, Zijian Yu, Huijia Zhu, Weiqiang Wang, Jun Wan, Zhen Lei
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 SFT (4K samples) to base model (ablation) | Qwen3-VL-8B · MintVid Others | accuracy: 55.2 -> 68.3 | Improved |
| add SFT (ablation) | Qwen3-VL-8B · MintVid Fact | accuracy: 91.9 -> 90.7 | Hurt |
| add joint DPO after SFT (ablation) | SFT model · MintVid Fact | accuracy: 90.7 -> 94.6 | Improved |
| add response-level DPO_t to J-DPO (ablation) | J-DPO · MintVid Others | accuracy: 66.3 -> 69.1 | Improved |
| add video-level DPO_v to J-DPO (ablation) | J-DPO · MintVid Fact | accuracy: 92.9 -> 94.6 | Improved |
| add J-DPO cold start before RL (vs pure RL) (ablation) | VideoVeritas · MintVid Others | accuracy: 69.1 -> 79.2 | Improved |
| add format reward to pure RL (ablation) | Pure RL · MintVid Fact | accuracy: 96.1 -> 87.5 | Hurt |
| add SSL object-counting perception task to PPRL (ablation) | VideoVeritas · MintVid Phantom-14B | accuracy: 77.1 -> 81.0 | Improved |
| add SSL + G-G perception tasks (ablation) | VideoVeritas · MintVid Seedance1.0Pro | accuracy: 81.3 -> 86.6 | Improved |
| add artifact grounding (A-G) on top of SSL + G-G (ablation) | VideoVeritas · MintVid Jimeng3.0Pro | accuracy: 80.8 -> 74.4 | Hurt |
| add 1K perception data (ablation) | PPRL · MintVid | accuracy: +1.1% Acc and 1.3% F1 | Improved |
| scale perception data beyond ~5K (ablation) | PPRL · MintVid | accuracy | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BusterX++ — BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation (NeurIPS 2026) · better on 4 of 4 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Video-R1: Reinforcing Video Reasoning in MLLMs (NeurIPS 2025 · similarity 0.88)
- Beyond Perceptual Shortcuts: Causal-Inspired Debiasing Optimization for Generalizable Video Reasoning in Lightweight MLLMs (CVPR 2026 · similarity 0.87)
- Fact-R1: Towards Explainable Video Misinformation Detection with Deep Reasoning (NeurIPS 2025 · similarity 0.87)
- NoisyRollout: Reinforcing Visual Reasoning with Data Augmentation (NeurIPS 2025 · similarity 0.87)
- When Thinking Drifts: Evidential Grounding for Robust Video Reasoning (NeurIPS 2025 · similarity 0.87)
- Incentivizing Versatile Video Reasoning in MLLMs via Data-Efficient Reinforcement Learning (CVPR 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.
