# Generative Video Synthesis (GVS)

[Back to benchmarks](README.md) · [Back to home](../README.md)

GVS benchmarks are the fastest-moving branch. They emphasize cross-generator transfer, physical plausibility, semantically fabricated scenarios, and robustness as generators rapidly improve.

| Date | Benchmark / Dataset | Paper | Focus | Venue |
| --- | --- | --- | --- | --- |
| 02/2026 | SynthForensics | [SynthForensics: A Multi-Generator Benchmark for Detecting Synthetic Video Deepfakes](https://doi.org/10.48550/ARXIV.2602.04939)<br>Leotta et al. | Human-centric synthetic-video benchmark with 6,815 videos from five open-source T2V generators. | arXiv |
| 02/2026 | MintVid | [VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning](https://doi.org/10.48550/ARXIV.2602.08828)<br>Tan et al. | Lightweight high-quality benchmark with 3K videos from nine state-of-the-art generators. | arXiv |
| 01/2026 | AIGVDBench | [Your One-Stop Solution for AI-Generated Video Detection](https://doi.org/10.48550/ARXIV.2601.11035)<br>Ma et al. | Large-scale benchmark with 440K videos from 31 generation models and 33 evaluated detectors. | arXiv |
| 12/2025 | ViFBench | [Skyra: AI-Generated Video Detection via Grounded Artifact Reasoning](https://arxiv.org/abs/2512.15693)<br>Li et al. | 3K videos from 10+ SOTA generators. | arXiv |
| 12/2025 | Video Reality Test | [Video Reality Test: Can AI-Generated ASMR Videos fool VLMs and Humans?](https://arxiv.org/abs/2512.13281)<br>Wang et al. | An ASMR-sourced benchmark. | arXiv |
| 10/2025 | ER-FF++set | [EDVD-LLaMA: Explainable Deepfake Video Detection via Multimodal Large Language Model Reasoning](https://arxiv.org/abs/2510.16442)<br>Sun et al. | Benchmark with supervision for detection + explanation. | arXiv |
| 10/2025 | AEGIS | [AEGIS: Authenticity Evaluation Benchmark for AI-Generated Video Sequences](https://doi.org/10.1145/3746027.3758295)<br>Li et al. | Large-scale benchmark for AIGC-V authenticity. | ACM MM |
| 09/2025 | DeeptraceReward | [Learning Human-Perceived Fakeness in AI-Generated Videos via Multimodal LLMs](https://arxiv.org/abs/2509.22646)<br>Fu et al. | Spatiotemporal benchmark with human-perceived fake traces. | arXiv |
| 07/2025 | GenBuster++ | [BusterX++: Towards Unified Cross-Modal AI-Generated Content Detection and Explanation with MLLM](https://arxiv.org/abs/2507.14632)<br>Wen et al. | A cross-modal benchmark for VLM evaluation. | arXiv |
| 06/2025 | GenWorld | [GenWorld: Towards Detecting AI-generated Real-world Simulation Videos](https://arxiv.org/abs/2506.10975)<br>Chen et al. | Large-scale real-world simulation dataset for AI-video detection. | arXiv |
| 06/2025 | Ivy-Fake | [Ivy-fake: A unified explainable framework and benchmark for image and video aigc detection](https://arxiv.org/abs/2506.00979)<br>Jiang et al. | Large-scale benchmark for explainable AIGC detection. | arXiv |
| 06/2025 | DAVID-X | [DAVID-XR1: Detecting AI-Generated Videos with Explainable Reasoning](https://arxiv.org/abs/2506.14827)<br>Gao et al. | AIGC-V with defect-level spatiotemporal annotations + rationales. | arXiv |
| 05/2025 | GenBuster-200K | [BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation](https://arxiv.org/abs/2505.12620)<br>Wen et al. | Real + synthetic videos simulating real-world conditions. | arXiv |
| 04/2025 | LOKI | [LOKI: A Comprehensive Synthetic Data Detection Benchmark using Large Multimodal Models](https://arxiv.org/abs/2410.09732)<br>Ye et al. | Multimodal synthetic-data detection benchmark over video, image, 3D, text, and audio. | ICLR |
| 03/2025 | Deepfake-Eval-2024 | [Deepfake-eval-2024: A multi-modal in-the-wild benchmark of deepfakes circulated in 2024](https://arxiv.org/abs/2503.02857)<br>Chandra et al. | Multi-modal in-the-wild benchmark of deepfakes from 2024. | arXiv |
| 01/2025 | GenVidBench | [Genvidbench: A challenging benchmark for detecting ai-generated video](https://arxiv.org/abs/2501.11340)<br>Ni et al. | AIGC-V detection dataset. | arXiv |
| 12/2024 | DVF | [On Learning Multi-Modal Forgery Representation for Diffusion Generated Video Detection](https://proceedings.neurips.cc/paper_files/paper/2024/file/dccbeb7a8df3065c4646928985edf435-Paper-Conference.pdf)<br>Song et al. | Diffusion Video Forensics dataset and benchmark. | NeurIPS |
| 05/2024 | GenVidDet | [Distinguish any fake videos: Unleashing the power of large-scale data and motion features](https://arxiv.org/abs/2405.15343)<br>Ji et al. | Real and AIGC-V videos from eight generation models. | arXiv |
| 05/2024 | GenVideo | [DeMamba: AI-Generated Video Detection on Million-Scale GenVideo Benchmark](https://arxiv.org/abs/2405.19707)<br>Chen et al. | AIGC-V detection dataset. | arXiv |
| 02/2024 | GVF | [Detecting AI-Generated Video via Frame Consistency](https://arxiv.org/abs/2402.02085)<br>Ma et al. | Generated Video Forensics benchmark for AI-video detection. | arXiv |