Research map / Video / Video segmentation and tracking
CoCoVideo: The High-Quality Commercial-Model-Based Contrastive Benchmark for AI-Generated Video Detection
Huidong Feng, Wentao Chen, Jie Chen, Xinqi Cai, Ruolong Ma, Yinglin Zheng, Yuxin Lin, Ming Zeng
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 paired contrastive projection head to the backbone (ablation) | CoCoDetect · CoCoVideo test ablation | accuracy (%): 81.05 -> 90.69 | Improved |
| add paired contrastive projection head to the backbone (ablation) | CoCoDetect · CoCoVideo test ablation | AUC (%): 91.39 -> 95.93 | Improved |
| add confidence-gated MLLM reasoning at inference (ablation) | CoCoDetect · CoCoVideo test ablation | accuracy (%): 88.92 -> 90.69 | Improved |
| add confidence-gated MLLM reasoning at inference (ablation) | CoCoDetect · CoCoVideo test ablation | AUC (%): 95.46 -> 95.93 | Improved |
| add contrastive training and confidence-gated MLLM to the 3D ResNet backbone (ablation) | 3D ResNet · CoCoVideo test ablation | accuracy (%): 78.69 -> 90.69 | Improved |
| add contrastive projection branch plus confidence gate to backbone (ablation) | 3D ResNet · CoCoVideo test ablation | F1 (%): 78.57 -> 90.62 | Improved |
| compare CoCoDetect with DeMamba on cross-dataset GVD | DeMamba · GVD (cross-dataset) | accuracy (%): 24.11 -> 71.49 | Improved |
| compare CoCoDetect with DeMamba on cross-dataset GVF | DeMamba · GVF (cross-dataset) | accuracy (%): 26.91 -> 74.56 | Improved |
| compare CoCoDetect with DeMamba on GenBuster | DeMamba · GenBuster (cross-dataset) | accuracy (%): 67.35 -> 64.80 | Hurt |
| compare CoCoDetect with DeMamba on GenVideo F1 | DeMamba · GenVideo (cross-dataset) | F1 (%): 57.31 -> 33.19 | Hurt |
| compare CoCoDetect with DeMamba on GenVideo accuracy | DeMamba · GenVideo (cross-dataset) | accuracy (%): 48.17 -> 57.04 | Improved |
| compare training-free D3 with trained methods | D3 · CoCoVideo test (average over 13 generators) | accuracy (%): 82.59 -> 48.95 | Hurt |
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.
- VideoMAE — VideoMAE: Masked Autoencoders are Data-Efficient Learners for Self-Supervised Video Pre-Training (NeurIPS 2022) · better on 2 of 2 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
- Forensics-Bench: A Comprehensive Forgery Detection Benchmark Suite for Large Vision Language Models (CVPR 2025 · similarity 0.89)
- BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation (NeurIPS 2026 · similarity 0.89)
- Generalizing Deepfake Video Detection with Plug-and-Play: Video-Level Blending and Spatiotemporal Adapter Tuning (CVPR 2025 · similarity 0.89)
- On Learning Multi-Modal Forgery Representation for Diffusion Generated Video Detection (NeurIPS 2024 · similarity 0.87)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.87)
- Exposing the Deception: Uncovering More Forgery Clues for Deepfake Detection (AAAI 2024 · similarity 0.87)
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.
