ant-research/Awesome-AIGC-Image-Video-Detection

A curated collection of the latest research and resources on AI-Generated Image and Video Detection.

459

52 commits

updated Sep 22, 2026

github.com/ant-research/Awesome-AIGC-Image-Video-Detection

See the code

README

Awesome AIGC Image/Video Detection

A curated collection of the latest research and resources on AI-Generated Image and Video Detection. This repository encompasses datasets, benchmarks, research papers, and practical detection tools.

🚀🚀🚀Contributions are welcome! If you find any missing papers, datasets, or tools, feel free to open an issue or submit a pull request.

Hot Events
Benchmarks & Datasets
Research Papers
- MLLM-Based
- Classification-Based
Competitions
Practical Detection Tools
About Our Team

🔥 Hot Events

Benchmarks & Datasets

Modality Legend: [I] Image | [V] Video | [M] Multi-modal

Annotation Type Legend: Au: Authenticity | Ex: Explainability | Lo: Localization

Benchmark	Paper	Venue & Year	Modality	Notes	Real Source	Fake Source/Generator	Annotation	Scale	Download
DF26	DF26: We Cannot Tell Fake From Real Anymore	Arxiv 2026	`[V]`	Modern-Generator AIGV Benchmark, Single-Person Public-Speaking Scenarios (Direct-to-Camera / Official Statement / Studio Interview), per-clip scene-prompt control (identity & scene fixed, generator varied), human + SOTA detector study near random chance, closed-source subset evaluation-only	OpenVid-1M, TalkingCelebs, MAVOS-DD	7 recent T2V/I2V models: Wan2.6, Veo 3.1, Grok Imagine 1.0, Kling 3.0 (closed); Wan2.2-A14B, HunyuanVideo 1.5, LTX 2.3 (open)	`Au`	2.7K (271 real + 2,420 fake, 1280×720)	DF26 (controlled access)
DailyBench	DailyBench: A Unified Benchmark for AI-Generated and Manipulated Images from Modern Generative Models	Arxiv 2026 (v3 2026-09-01)	`[I]`	Unified AIID testbed = FakeBench (full T2I synthesis) + ManipulationBench (object-level edits on real images); LAION-Aesthetics V2 pool filtered by aesthetic score ≥6 & shortest side ≥512; per-generator real/fake pairs; robustness study under recompression & pixel perturbation; ships FPD diagnostic baseline	LAION-Aesthetics V2	T2I: SD3.5-Large, FLUX.1, FLUX.2, Qwen-Image-2512, Z-Image, Nano Banana 2, GPT-Image 2; Edit: FLUX-Fill (random/object mask), FLUX.2-klein-9B, Qwen-Image-Edit-2511, Step1X-Edit-v1p2	`Au`	270K（v3：FakeBench ≈185K + ManipulationBench ≈75K）	DailyBench Project
AGIDefect-4K	AGIDefect-4K: A Richly Annotated Dataset for AI-Generated Image Defect Detection, Localization and Explanation	ACM MM 2026	`[I]`	Defect Detection, Localization & Explanation, 15 SOTA Generators, Quality Scoring		DALL-E 3, Midjourney, FLUX, Gemini, GPT-Image, Ideogram, Kling, Grok, etc.	`Au`, `Lo`, `Ex`	4K	AGIDefect-4K
RA-Bench	Can We Defend Against AI-Generated Video Attacks on Real-World Crisis Events? A Systematic Evaluation of Detectors, Generators and Social Dissemination	Arxiv 2026	`[V]`	Real-Crisis-Anchored, Source-Matched Evaluation, Human-Proof Subset, Propagation Robustness	Real crisis event footage (public media & source URLs)	4 open-source + 5 closed-source generators (incl. Wan2.2)	`Au`	17.9K	RA-Bench
RealHD	RealHD: A High-Quality Dataset for Robust Detection of State-of-the-Art AI-Generated Images	Arxiv 2026	`[I]`	Multi-category, SOTA Generators, 10K+ Prompts, Inpainting Masks		T2I, Inpainting, Refinement, Face Swapping	`Au`, `Lo`	730K+	RealHD
Treasure	Fleet: Few Shots Lead Effective AI-generated Image Detection	ICML 2026	`[I]`	64 Models, 20 Closed-source Commercial Engines, Few-shot Adaptation		Diverse architectures & 20 commercial engines	`Au`	360K	Treasure
LADBench	LADBench: A Benchmark for Logical Fault Detection in Images	ICDL 2026	`[I]`	Logical Anomaly Detection, VLM Evaluation, Common Sense Reasoning		Synthetic images with logical anomalies	`Au`	1K+	LADBench
EVID-Bench	When Seeing Is Not Believing -- A Benchmark for Search-Grounded Video Misinformation Detection	Arxiv 2026	`[V]`	Search-Grounded Verification, Evidence-Dependent Manipulation			`Au`		EVID-Bench
CoCoVideo	CoCoVideo: The High-Quality Commercial-Model-Based Contrastive Benchmark for AI-Generated Video Detection	Arxiv 2026	`[V]`	Commercial AIGC Models, Contrastive Benchmark		Commercial video generation models	`Au`		CoCoVideo
FraudBench	FraudBench: A Multimodal Benchmark for Detecting AI-Generated Fraudulent Refund Evidence	Arxiv 2026	`[M]`	Fraudulent Refund Detection (ECommerce)			`Au`		FraudBench
GPT-Image-2 Wild	GPT-Image-2 in the Wild: A Twitter Dataset of Self-Reported AI-Generated Images from the First Week of Deployment	Arxiv 2026	`[I]`	GPT-Image-2	Twitter (real images)	GPT-Image-2	`Au`	10K	GPT-Image-2 Wild
Artifact-Bench	Artifact-Bench: Evaluating MLLMs on Detecting and Assessing the Artifacts of AI-Generated Videos	Arxiv 2026	`[V]`	MLLM Evaluation, Video Artifacts			`Au`		Artifact-Bench
CommGen15	PGC: Peak-Guided Calibration for Generalizable AI-Generated Image Detection	ICML 2026	`[I]`	15 Commercial Generative Models		15 commercial generators	`Au`		CommGen15
AEGIS-Academic	AEGIS: A Holistic Benchmark for Evaluating Forensic Analysis of AI-Generated Academic Images	Arxiv 2026	`[I]`	Academic Image Forensics			`Au`		AEGIS
SciFigDetect	SciFigDetect: A Benchmark for AI-Generated Scientific Figure Detection	Arxiv 2026	`[I]`	Scientific Figure Detection		Nano Banana Pro, GPT-image-1.5	`Au`	150K	SciFigDetect
ActivityForensics	ActivityForensics: A Comprehensive Benchmark for Localizing Manipulated Activity in Videos	CVPR 2026	`[V]`	Action-level AIGC in videos			`Au`	6K	ActivityForensics
MintVid	VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning	ICML 2026	`[V]`		OpenVid, VFHQ, HDTF, TikTok	Jimeng3.0-Pro, Seedance, Kling2.5-Turbo, Sora2, TikTok, Youtube, etc.	`Au`	4K	MintVid
AIGVDBench	Your One-Stop Solution for AI-Generated Video Detection	CVPR 2026	`[V]`		OpenVid-HD	31 generation models	`Au`	440k	AIGVDBench
HydraFake	Veritas: Generalizable Deepfake Detection via Pattern-Aware Reasoning	ICLR 2026(Oral)	`[I]`		FFHQ, VFHQ, CelebAHQ, FF++, etc.	GPT-4o, HailuoAI, ICLight, InfiniteYou, etc.	`Au`, `Ex`	100K	HydraFake
BR-Gen	Zooming In on Fakes: A Novel Dataset for Localized AI-Generated Image Detection with Forgery Amplification Approach	AAAI 2026	`[I]`				`Au`, `Lo`	150K	BR-Gen
RRDataset	Bridging the Gap Between Ideal and Real-world Evaluation: Benchmarking AI-Generated Image Detection in Challenging Scenarios	ICCV 2025	`[I]`	Real-World Robustness, Internet Transmission, Re-digitization			`Au`		DownLoad
HiResolution	No Pixel Left Behind: A Detail-Preserving Architecture for Robust High-Resolution AI-Generated Image Detection	ICLR 2026	`[I]`				`Au`	50K	HiRes-50K
AIGI-Now	AlignGemini: Generalizable AI-Generated Image Detection Through Task-Model Alignment	Arxiv 2026	`[I]`		COCO	Nano Banana, GPT-4o, Jimeng, Kling, Minimax, etc.	`Au`	18K	AIGI-Now
RealChain	Beyond Artifacts: Real-Centric Envelope Modeling for Reliable AI-Generated Image Detection	Arxiv 2026	`[I]`				`Au`	14K	RealChain
GenVidBench	GenVidBench: A 6-Million Benchmark for AI-Generated Video Detection	AAAI 2026	`[V]`				`Au`	6M	GenVidBench
Skyra	Skyra: AI-Generated Video Detection via Grounded Artifact Reasoning	CVPR 2026	`[V]`				`Au`, `Ex`, `Lo`	4K	ViF-CoT-4K
So-Fake-Set	So-Fake: Benchmarking and Explaining Social Media Image Forgery Detection	Arxiv 2025	`[I]`		F30k, WIDER, FFHQ, CelebA, OpenImages, COCO, OpenForensics	Qwen-image, GPT-4o, Nano Banana, Seedream3.0, Ideogram3.0, etc.	`Au`	2M+	So-Fake-Set So-Fake-OOD
GenBuster++	BusterX++: Towards Unified Cross-Modal AI-Generated Content Detection and Explanation with MLLM	Arxiv 2025	`[M]`				`Au`	4K	GenBuster++
GenBuster	BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation	Arxiv 2025	`[I]`				`Au`	200K	GenBuster-200K
AIGIBench	Is Artificial Intelligence Generated Image Detection a Solved Problem?	NeurIPS 2025	`[I]`		FFHQ, CelebA-HQ, Open Images V7	Common generators & SocialRF, CommunityAI	`Au`	200K	AIGIBench
Ivy-Fake	IVY-FAKE: A Unified Explainable Framework and Benchmark for Image and Video AIGC Detection	Arxiv 2025	`[M]`				`Au`, `Ex`	150K	Ivy-Fake
AEGIS	AEGIS: Authenticity Evaluation Benchmark for AI-Generated Video Sequences	ACM MM 2025	`[V]`		Vript (YouTube, TikTok), DVF, YouTube (self-collected)	Stable Video Diffusion, CogVideoX-5B, I2VGen-XL, Pika, KLing, Sora	`Au`, `Ex`	10K+	AEGIS
NeXT-IMDL	NeXT-IMDL: Build Benchmark for Next-Generation Image Manipulation Detection & Localization	Arxiv 2025	`[I]`		Flickr30k, COCO, OpenImages V7	SD2-Inpainting, SDXL-Inpainting, FLUX-Inpainting, etc.	`Au`, `Lo`	558K	NeXT-IMDL
ARForensics	D3QE: Learning Discrete Distribution Discrepancy-aware Quantization Error for Autoregressive-Generated Image Detection	ICCV 2025	`[I]`		ImageNet	Infinity, Janus_Pro, RAR, Switti, VAR, LlamaGen, Open_MAGVIT2	`Au`	300k	ARForensics
OpenSDI	OpenSDI: Spotting Diffusion-Generated Images in the Open World	CVPR 2025	`[I]`		Megalith-10M	SD1.5, SD2.1, SDXL, SD3, Flux.1	`Au`, `Lo`	300K	OpenSDI
Community Forensics	Community Forensics: Using Thousands of Generators to Train Fake Image Detectors	CVPR 2025	`[I]`		LAION, ImageNet, COCO, FFHQ, CelebA, MetFaces, AFHQ, etc.	4803 generators (Latent Diffusion, GAN, Autoregressive, Pixel Diffusion, Commercial)	`Au`	2.7M	Community Forensics
FakeClue	Spot the Fake: Large Multimodal Model-Based Synthetic Image Detection with Artifact Explanation	NeurIPS 2025	`[I]`				`Au`, `Ex`	100K	FakeClue
XAIGID-RewardBench	Explainable AI-Generated Image Detection RewardBench	NeurIPS 2025 Workshop	`[I]`		COCO-2017	Imagen 4, Flux.1 Dev, Bagel, etc.	`Au`, `Ex`	3K	XAIGID-RewardBench
RewardData	Learning Human-Perceived Fakeness in AI-Generated Videos via Multimodal LLMs	Arxiv 2025	`[V]`				`Au`, `Ex`	4.3K	RewardData
OpenFake	OPENFAKE: An Open Dataset and Platform Toward Real-World Deepfake Detection	Arxiv 2025	`[I]`		LAION-400M	SD 1.5/2.1/XL/3.5, Flux 1.0-dev/1.1-Pro/Schnell, Midjourney v6/v7, DALL·E 3, Imagen 3/4, GPT Image 1, Ideogram 3.0, Grok-2, HiDream-I1, Recraft v3, Chroma, and 10 community LoRA/finetune variants	`Au`	~4M	OPENFAKE
Video Reality Test	Video Reality Test: Can AI-Generated ASMR Videos fool VLMs and Humans?	Arxiv 2025	`[V]`		YouTube ASMR (social media)	Veo3.1-Fast, Sora2, Wan2.2-A14B, Wan2.2-5B, OpenSora-V2, HunyuanVideo, StepVideo	`Au`	149 real + dynamic fake	Video Reality Test
DDL	DDL: A Dataset for Interpretable Deepfake Detection and Localization in Real-World Scenarios	Arxiv 2025	`[M]`				`Au`	367K	DDL
DiffSeg30k	DiffSeg30k: A Multi-Turn Diffusion Editing Benchmark for Localized AIGC Detection	Arxiv 2025	`[I]`		COCO	SD2, SD3.5, SDXL, Flux.1, Glide, Kolors, HunyuanDiT1.1, Kandinsky 2.2	`Au`, `Lo`	30K	DiffSeg30k
FakeParts	FakeParts: a New Family of AI-Generated DeepFakes	Arxiv 2025	`[V]`				`Au`, `Lo`	81K	FakeParts
ForensicHub	ForensicHub: A Unified Benchmark & Codebase for All-Domain Fake Image Detection and Localization	NeurIPS 2025	`[I]`			ProGAN, StyleGAN, LDM, SDv1.4, SDv1.5, SDv2, SDXL, SD-ControlNet, MidJourney, ADM, GLIDE, VQDM, BigGAN	`Au`, `Lo`	23 datasets 42 models	ForensicHub
LOKI	LOKI: A Comprehensive Synthetic Data Detection Benchmark Using Large Multimodal Models	ICLR 2025	`[M]`			SORA, Keling, Open-Sora, FLUX, Midjourney, Stable Diffusion, Nerf-based, Gaussian-based, GPT-4o, Qwen-Max, Llama 3.1-405B, MusicGen, AudioLDM2...	`Au`, `Ex`	18K	LOKI
Chameleon	A Sanity Check for AI-Generated Image Detection	ICLR 2025	`[I]`		Unsplash	Midjourney, DALLE-3, Stable Diffusion (various LoRA fine-tuned)	`Au`	26K	Chameleon
WildFake	WildFake: A Large-scale Challenging Dataset for AI-Generated Images Detection	AAAI 2025	`[I]`				`Au`	3.7M	WildFake
WildRF	Real-Time Deepfake Detection in the Real-World	Arxiv 2024	`[I]`		Reddit, X (Twitter), Facebook (real images)	Reddit, X (Twitter), Facebook (social media deepfakes)	`Au`		WidlRF
AIGCDetectBenchmark	PatchCraft: Exploring Texture Patch for Efficient AI-generated Image Detection	Arxiv 2024	`[I]`				`Au`	100K	AIGCDetectionBenchMark
GenVideo	DeMamba: AI-Generated Video Detection on Million-Scale GenVideo Benchmark	Arxiv 2024	`[V]`				`Au`	2.3M	GenVideo
DRCT	Drct: Diffusion reconstruction contrastive training towards universal detection of diffusion generated images	ICML 2024	`[I]`		MSCOCO	LDM, SDv1.4, SDv1.5, SDv2, SDXL, SD-ControlNet	`Au`	2M	DRCT-2M
GenImage	GenImage: A Million-Scale Benchmark for Detecting AI-Generated Image	NeurIPS 2023	`[I]`		ImageNet, Wukong	MidJourney, SDv1.4, SDv1.5, ADM, GLIDE, VQDM, BigGAN	`Au`	2.7M	GenImage
DF40	DF40: Toward Next-Generation Deepfake Detection	NeurIPS 2024	`[I]` `[V]`				`Au`	0.1M+ videos, 1M+ images	DF40
Forensics-Bench	Forensics-Bench: A Comprehensive Forgery Detection Benchmark Suite for Large Vision Language Models	CVPR 2025	`[I]`, `[V]`, `[M]`		Various public datasets	GAN, Diffusion, VAE, RNN, Encoder-Decoder, Graphics-based	`Au`, `Lo`	63K	Forensics-Bench

⬆ Back to Top

Research Papers

💡 Note: Papers are sorted by year (descending) within each category.
Modality Legend: [I] Image | [V] Video | [M] Multi-modal
Category Layout: Top level is split into MLLM-Based (MLLM-powered detection) and Classification-Based (compact/traditional classifiers). Classification-Based is further organized into six subcategories. When a paper fits multiple subcategories, the priority is: Training-Free/Zero-Shot > Continual/Incremental > Video Spatiotemporal > Frequency/Low-Level Artifacts > Supervised General.

MLLM-Based

This category focuses on utilizing Multimodal Large Language Models (MLLMs) like GPT-4V, LLaVA, or Qwen-VL to detect AI-generated content. These methods often provide natural language explanations (explainability) alongside binary detection.

Title	Venue & Year	Modality	Highlights/Keywords	Code
Evidence-Guided Detection, Localization and Explanation for Text-Centric Image Forensics	ACM MM 2026 Challenge	`[I]`	Detector-Localizer-Reasoner Cascade, Iterative Difficulty-Aware Mining, Report-Mask Consistency	GitHub
AGIDefect-4K: A Richly Annotated Dataset for AI-Generated Image Defect Detection, Localization and Explanation	ACM MM 2026	`[I]`	AGIDefect-4K Dataset, Hierarchical Defect Annotation, MLLM Baseline (AGIDA)	GitHub
Explainable Deepfake Detection with Feature-robust Augmentation and Evidence-grounded Explanation Optimization	ACM MM 2026	`[I]`	Feature-robust Augmentation, Mean-Teacher Consistency, Evidence-grounded Preference Optimization, Challenge Winner	GitHub
PATE-Forensics: Perception-as-Tool for Explainable Deepfake Forensics with General-Purpose MLLMs	IJCAI 2026 Workshop	`[I]`	Perception-as-Tool, DINOv3 Forensic Perception Tool, General-Purpose MLLM Explanation	GitHub
Defake-o3: From Speculative Rationales to Verifiable Evidence for Explainable AIGI Detection	ACM MM 2026	`[I]`	Interactive Visual Search, Verifier-guided Evidence Alignment, GroundFake Dataset, FakeFrontier Benchmark	N/A
SPARED: Reasoning-Based AI-Generated Image Detection via Adversarially Edited Data	Arxiv 2026	`[I]`	Adversarial RL Loop, Diffusion Editor, Free-form Explanation	N/A
VidForensics-M1: Meta-Detection Reinforcement Learning with Verifiable Temporal Grounding for AI-Generated Video Forensics	Arxiv 2026	`[V]`	Meta-Detection RL, Verifiable Temporal Grounding, Evidence-Guided Reward Redistribution	N/A
Veritas++: Value-aware On-Policy Distillation for Perception-Enhanced AIGI Detection	Arxiv 2026	`[I]`	Value-aware On-Policy Distillation, Perception-Enhanced Reasoning	GitHub
Detecting AI-Generated Video: A Vision-Language Dual-View Survey	ACL 2026 Findings	`[V]`	[Survey] Vision-Language Dual-View Taxonomy, Factual Fidelity Verification, Cross-modal Consistency	N/A
TranX-Adapter: Bridging Artifacts and Semantics within MLLMs for Robust AI-generated Image Detection	ICML 2026	`[I]`	Artifact feature and Semantic feature fusion	GitHub
Venus-DeFakerOne: Unified Fake Image Detection & Localization	Arxiv 2026	`[I]`	Unified Detection & Localization, Large-Scale Training	GitHub
GenShield: Unified Detection and Artifact Correction for AI-Generated Images	ICML 2026	`[I]`	Unified MLLM, Detect & Correct Artifacts	GitHub
ReAlign: Generalizable Image Forgery Detection via Reasoning-Aligned Representation	Arxiv 2026	`[I]`	Reasoning-Aligned Representation, Interpretable	N/A
UniGenDet: A Unified Generative-Discriminative Framework for Co-Evolutionary Image Generation and Generated Image Detection	CVPR 2026	`[I]`	Unified Model, Co-Evolution (Generation & Detection)	GitHub
VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning	ICML 2026	`[V]`	Perception Pretext RL, Fact-based Reasoning, MintVid Dataset	GitHub
Veritas: Generalizable deepfake detection via pattern-aware reasoning	ICLR 2026(Oral)	`[I]`	Pattern-aware Reasoning, HydraFake Dataset	Github
DF-LLaVA: Unlocking MLLMs for Synthetic Image Detection via Knowledge Injection and Conflict-Driven Self-Reflection	Arxiv 2026	`[I]`	Knowledge Injection, Self-Reflection	N/A
DocShield: Towards AI Document Safety via Evidence-Grounded Agentic Reasoning	Arxiv 2026	`[M]`	Agentic Framework, Document Safety	N/A
VidGuard-R1: AI-Generated Video Detection and Explanation via Reasoning MLLMs and RL	ICLR 2026	`[V]`	Multi-stage RL (GRPO), Time Artifacts, Video Detection Dataset	GitHub
FakeXplain: AI-Generated Image Detection via Human-Aligned Grounded Reasoning	ICLR 2026	`[I]`	Grounded Reasoning, Human-annotated Dataset	N/A
AlignGemini: Generalizable AI-Generated Image Detection Through Task-Model Alignment	Arxiv 2026	`[I]`	Decoupling (Semantic & Pixel), AIGI-Now Dataset	N/A
Zoom-In to Sort AI-Generated Images Out	Arxiv 2026	`[I]`	Thinking with Images, MagniFake Dataset	N/A
AgentFoX: LLM Agent-Guided Fusion with eXplainability for AI-Generated Image Detection	Arxiv 2026	`[I]`	Agentic framework	Github
EvoGuard: An Extensible Agentic RL-based Framework for Practical and Evolving AI-Generated Image Detection	Arxiv 2026	`[I]`	Agentic Framework, Method Ensembling	N/A
VIGIL: Part-Grounded Structured Reasoning for Generalizable Deepfake Detection	Arxiv 2026	`[I]`	Part-centric Forensic, OmniFake Dataset	Project
GenVideoLens: Where LVLMs Fall Short in AI-Generated Video Detection?	Arxiv 2026	`[V]`	GenVideoLens benchmark	N/A
Semantic Visual Anomaly Detection and Reasoning in AI-Generated Images	ICLR 2026	`[I]`	Semantic Anomaly Reasoning, AnomReason Dataset	N/A
FAKE-HR1: RETHINKING REASONING OF VISION LANGUAGE MODEL FOR SYNTHETIC IMAGE DETECTION	Arxiv 2026	`[I]`	Hybrid-Reasoning, Dual-mode Dataset	N/A
MIRAGE: Towards AI-Generated Image Detection in the Wild	Arxiv 2025	`[I]`	Human Curation Dataset, Heuristic-to-Analytic Reasoning	N/A
BusterX++: Towards Unified Cross-Modal AI-Generated Content Detection and Explanation with MLLM	Arxiv 2025	`[M]`	RL Post-training, Cross-Modal, Thinking Reward Mechanism	Github
BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation	Arxiv 2025	`[V]`	GenBuster-200K Dataset, Cold Start + RL Training	Github
REVEAL: Reasoning-enhanced Forensic Evidence Analysis for Explainable AI-generated Image Detection	Arxiv 2025	`[I]`	Chain-of-Evidence, Expert-grounded RL	N/A
Spot the Fake: Large Multimodal Model-Based Synthetic Image Detection with Artifact Explanation	NeurIPS 2025	`[I]`	FakeClue Dataset, Fine-grained Artifact Clues, Artifact Explanation	GitHub
AIGI-Holmes: Towards Explainable and Generalizable AI-Generated Image Detection via Multimodal Large Language Models	ICCV 2025	`[I]`	Holmes-Set, Multi-Expert Jury, 3-Stage Training Pipeline	Github
LEGION: Learning to Ground and Explain for Synthetic Image Detection	ICCV 2025	`[I]`	SynthScars Dataset, Defender & Controller, Image Refinement	GitHub
Seeing Before Reasoning: A Unified Framework for Generalizable and Explainable Fake Image Detection	Arxiv 2025	`[I]`	Perception & Reasoning, ExplainFake-Bench	N/A
SIDA: Social Media Image Deepfake Detection, Localization, and Explanation	CVPR 2025	`[I]`	SID-Set, Mask Prediction, Social Media Context	Github
FakeShield: Explainable Image Forgery Detection and Localization via Multi-modal Large Language Models	ICLR 2025	`[I]`	Explainable IFDL, Domain Tag-guided, Multi-modal Localization	GitHub
FakeScope: Large Multimodal Expert Model for Transparent AI-Generated Image Forensics	Arxiv 2025	`[I]`	FakeChain Dataset, FakeInstruct, Trace Evidence	N/A
AntifakePrompt: Prompt-Tuned Vision-Language Models are Fake Image Detectors	Arxiv 2024	`[I]`	VQA, InstructBLIP, Soft Prompt-tuning, Zero-shot	GitHub

Classification-Based

This category includes supervised learning approaches that train neural networks (CNNs, ViTs, VFMs, etc.) specifically to classify authentic vs. AI-generated content. They usually focus on robustness, generalization, and feature extraction. It is organized into six subcategories: Supervised General Detectors, Video Spatiotemporal Modeling, Training-Free / Zero-Shot, Frequency-Domain & Low-Level Artifacts, Continual & Incremental Learning, and Related & Other.

Supervised General Detectors

Trainable classifiers and backbones (CNNs, ViTs, vision foundation models, CLIP-based adapters, few-shot and prompt-based methods) for general-purpose detection.

Title	Venue & Year	Modality	Highlights/Keywords	Code
Learning Continuous Source Responses For Generalizable AI-Generated Image Detection	Arxiv 2026	`[I]`	CuRe, Continuous Source-Response Regression (Real-Generated Mixing Ratio), Shortcut-Cue Suppression, Source-Response Subspace, 10-Benchmark Evaluation	GitHub
Harnessing CLIP and DINO: An Uncertainty-Aware Cascaded Fusion Network for Generalizable Deepfake Image Detection	Arxiv 2026	`[I]`	UCF-Net, CLIP Semantic + DINO Structural Priors, Layer-wise Expert Aggregation, Entropy-based Uncertainty Fusion, 4M-image Unified Benchmark, Cross-generator Evaluation	GitHub
FUSED: Forensic-Semantic Mixture-of-Experts for AI Inpainting Detection and Localization	Arxiv 2026	`[I]`	Forensic-Semantic MoE, Joint Detection & Localization, Cross-generator Generalization (OpenSDID)	GitHub
GAP-SAM: A Global Artifact Prior for Generalizable AI-Generated Image Manipulation Localization	Arxiv 2026	`[I]`	Global Artifact Token, SAM3 FiLM Injection, Boundary Adhesion Analysis	N/A
LoRC: Detecting AI-Generated Images via Low-Rank Collapse in Semantic Residuals	ECCV 2026 (Spotlight)	`[I]`	Low-Rank Collapse Signature, Semantic-Residual Decoupling, Cross-model Generalization	N/A
Prior-Conditioned Gaussian Discriminants for Generalizable AI-generated Image Detection	ECCV 2026	`[I]`	Closed-form Gaussian Heads, Percept-Lens Protocol (39 Datasets), Transfer Diagnostic	N/A
Environment-Invariant Subspace Learning for Generalizable Deepfake Detection	Arxiv 2026	`[I]`	Environment-Invariant Subspace, VFM Semantic Priors, Environmental Intervention	N/A
Understanding Why Foundation Models Work for Diffusion-Generated Image Detection	Arxiv 2026	`[I]`	Interpretability Analysis, DDIM Inversion, Low-to-Mid Frequency Distributional Discrepancy	N/A
PatchHead: Learning Spatial Patch Evidence for Generalizable AI-Generated Image Detection	Arxiv 2026	`[I]`	DINO Patch Tokens, 2D Spatial Aggregation, LoRA Adapters	N/A
GlobalForge: Towards Robust AI-Generated Image Detection	Arxiv 2026	`[I]`	Global Structural Reasoning, Local Information Bottleneck, RealDeg-Bench	Code
Fleet: Few Shots Lead Effective AI-generated Image Detection	ICML 2026	`[I]`	Few-shot Adaptation, Routing Correction, Treasure Benchmark	GitHub
SSAFE: Simple and Strong AI-Generated Image Detection via Frozen Vision Encoders	Arxiv 2026	`[I]`	Frozen Vision Encoders, Linear Classifier, RealWorldBench	N/A
HydraPrompt: An Adaptive and Asymmetric Framework of Vision-Language Models for Synthetic Image Detection	ACM MM 2026	`[I]`	Asymmetric Prompting, Dynamic Decision Boundary	N/A
VINA: Video as Natural Augmentation: Towards Unified AI-Generated Image and Video Detection	Arxiv 2026	`[M]`	Unified Image/Video Detection, Cross-Modal Contrastive Learning	N/A
PGC: Peak-Guided Calibration for Generalizable AI-Generated Image Detection	ICML 2026	`[I]`	Peak-Guided Calibration, CommGen15 Dataset	GitHub
Reduce the Artifacts Bias for More Generalizable AI-Generated Image Detection	Arxiv 2026	`[I]`	Bias-free Training	GitHub
Zooming In on Fakes: A Novel Dataset for Localized AI-Generated Image Detection with Forgery Amplification Approach	AAAI 2026	`[I]`	Localized AIGC Detection, Forgery Amplification, Scene-aware Local Forgery	GitHub
Simplicity Prevails: The Emergence of Generalizable AIGI Detection in Visual Foundation Models	Arxiv 2026	`[I]`	Linear Probe, Vision Foundation Models, Emergent Forensic Capability	N/A
MIRROR: Manifold Ideal Reference ReconstructOR for Generalizable AI-Generated Image Detection	Arxiv 2026	`[I]`	Manifold Reconstruction, Memory Bank, Human-AIGI Benchmark	GitHub
No Pixel Left Behind: A Detail-Preserving Architecture for Robust High-Resolution AI-Generated Image Detection	ICLR 2026	`[I]`	Detail-preserving dual-path architecture, Multi-task learning, HiRes-50K benchmark	N/A
All Patches Matter, More Patches Better: Enhance AI-Generated Image Detection via Panoptic Patch Learning	ICLR 2026	`[I]`	Random Patch Replacement, Patch-wise Contrastive Learning	N/A
OmniAID: Decoupling Semantic and Artifacts for Universal AI-Generated Image Detection in the Wild	Arxiv 2025	`[I]`	Mixture-of-Experts, Semantic-Artifact Decoupling, Mirage Dataset	GitHub
DINO-Detect: A Simple yet Effective Framework for Blur-Robust AI-Generated Image Detection	Arxiv 2025	`[I]`	Blur Robustness, Knowledge Distillation, DINOv3	Github
Orthogonal Subspace Decomposition for Generalizable AI-Generated Image Detection	ICML 2025 (Oral)	`[I]`	SVD Orthogonal Subspace, Asymmetry Phenomenon, Parameter-efficient Fine-tuning	GitHub
A Bias-Free Training Paradigm for More General AI-generated Image Detection	CVPR 2025	`[I]`	Bias-Free, Semantic Alignment, Stable Diffusion Self-conditioning	Github
Forensics Adapter: Adapting CLIP for Generalizable Face Forgery Detection	CVPR 2025	`[I]`	CLIP, Blending Boundaries, Forgery-aware Prompt Learning	Github
Exploring Unbiased Deepfake Detection via Token-Level Shuffling and Mixing	AAAI 2025	`[I]`	Token-Level Shuffling, Contrastive Loss, Bias Mitigation	N/A
FakeFormer: Efficient Vulnerability-Driven Transformers for Generalisable Deepfake Detection	Arxiv 2024	`[I]`	Vulnerability-driven, Local Attention (L2-Att), Vision Transformer	GitHub

Video Spatiotemporal Modeling

Methods that exploit temporal inconsistencies, motion patterns, and spatiotemporal artifacts in AI-generated videos.

Title	Venue & Year	Modality	Highlights/Keywords	Code
Mind the Rift: Cross-Scale Coupling Mismatch for AI-Generated Video Detection	ACM MM 2026	`[V]`	Cross-Scale Coupling Mismatch, Macro Temporal Dynamics vs. Pixel-Level Residuals, Persistent Homology, Encoder-Agnostic	GitHub
MotionPhys: Detecting AI-Generated Videos via Physical Consistency of Optical-Flow Trajectories	Arxiv 2026	`[V]`	Physical Motion Consistency, Sparse Optical-Flow Trajectories, Multi-scale Geometric Evolution	N/A
Rethinking the Readout: Unlocking Video Backbones for AI-Generated Video Detection	Arxiv 2026	`[V]`	V-PVP Readout, Patch Velocity Profiling, Frozen Video Backbones	Code
Dataset Biases and Shortcut Learning in Motion-Based AI-Generated Video Detection	Arxiv 2026	`[V]`	Motion Bias Analysis, Preprocessing/Sampling Bias, Frequency-based Comparison	N/A
G2VD: Generalizable AI-Generated Video Detection via Counterfactual Intervention and Causal Disentanglement	Arxiv 2026	`[V]`	Counterfactual Intervention, Causal Disentanglement, Cross-domain Generalization	GitHub
ReConFuse: Reconstruction-Error Guided Semantic Fusion for AI-Generated Video Detection	Arxiv 2026	`[V]`	Reconstruction Error, Semantic Fusion, Spatial-Temporal Artifacts	N/A
Detecting AI-Generated Videos with Spiking Neural Networks	Arxiv 2026	`[V]`	Spiking Neural Networks, Temporal Artifact	N/A
CMTA: Leveraging Cross-Modal Temporal Artifacts for Generalizable AI-Generated Video Detection	Arxiv 2026	`[V]`	Cross-Modal Temporal Artifacts, Video Detection	N/A
Preserving Forgery Artifacts: AI-Generated Video Detection at Native Scale	ICLR 2026	`[V]`	Native scale video processing, Massive realistic video dataset, Preserves subtle generation artifacts	N/A
Seeing What Matters: Generalizable AI-generated Video Detection with Forensic-Oriented Augmentation	NeurIPS 2025	`[V]`	Wavelet-band Augmentation, Forensic Frequency Artifacts, Single-generator Generalization	GitHub
AI-Generated Video Detection via Perceptual Straightening	NeurIPS 2025	`[V]`	Perceptual Straightening, DINOv2, Temporal Curvature	GitHub
Physics-Driven Spatiotemporal Modeling for AI-Generated Video Detection	NeurIPS 2025	`[V]`	Normalized Spatiotemporal Gradient (NSG), Maximum Mean Discrepancy (MMD)	Github
Towards a Universal Synthetic Video Detector: From Face or Background Manipulations to Fully AI-Generated Content	CVPR 2025	`[V]`	SigLIP-So400M, Attention-Diversity Loss, Full-frame Manipulations	N/A
DIP: Diffusion Learning of Inconsistency Pattern for General DeepFake Detection	TMM 2025	`[V]`	Direction-aware Attention, SpatioTemporal Invariant Loss	N/A
DeMamba: AI-Generated Video Detection on Million-Scale GenVideo Benchmark	Arxiv 2024	`[V]`	Mamba, State Space Model, Long-range Spatiotemporal Inconsistency	GitHub
Distinguish Any Fake Videos: Unleashing the Power of Large-scale Data and Motion Features	Arxiv 2024	`[V]`	GenVidDet, Optical Flow, Dual-Branch 3D Transformer	N/A

Training-Free / Zero-Shot

Methods that detect AI-generated content without additional training on detection data.

Title	Venue & Year	Modality	Highlights/Keywords	Code
Frozen DINO Localizes Image Edits Without a Localizer	Arxiv 2026	`[I]`	Training-free, Frozen DINO Patch-token Drift, Haar Perturbation, Edit Localization	GitHub
SPLIT: Training-Free AI-Generated and Partially Edited Video Detection via Spatial Patch-Level Incoherence and Temporal Roughness	ECCV 2026	`[V]`	Training-free, Patch-level Incoherence, Temporal Roughness, Ultra-low FPR	GitHub
Training-free Detection of Generated Videos via Spatial-Temporal Likelihoods	CVPR 2026	`[V]`	Training-free, Zero-shot, Spatial-Temporal Likelihoods, ComGenVid Dataset	GitHub

Frequency-Domain & Low-Level Artifacts

Methods based on spectral analysis, quantization/upsampling traces, and other low-level generative artifacts.

Title	Venue & Year	Modality	Highlights/Keywords	Code
Structured Local Differential Modeling for AI-Generated Image Detection	Arxiv 2026	`[I]`	RippleNet, Local Differential Signals, Low-SNR Forgery Traces	N/A
Dual Data Alignment Makes AI-Generated Image Detector Easier Generalizable	NeurIPS 2025 (Spotlight)	`[I]`	Dual-domain Alignment, Frequency-level Bias, VAE Reconstruction	GitHub
D3QE: Learning Discrete Distribution Discrepancy-aware Quantization Error for Autoregressive-Generated Image Detection	ICCV 2025	`[I]`	Discrete Distribution Discrepancy-aware Transformer, Vector Quantized Variational AutoEncoder	Github
Any-Resolution AI-Generated Image Detection by Spectral Learning	CVPR 2025	`[I]`	Spectral Context Attention, Frequency Reconstruction, OOD Detection	Github
Frequency-Aware Deepfake Detection: Improving Generalizability through Frequency Space Domain Learning	AAAI 2024	`[I]`	Frequency Domain, FFT, Frequency Conv Layer (FCL), Lightweight	GitHub
Rethinking the Up-Sampling Operations in CNN-based Generative Network	CVPR 2024	`[I]`	Neighboring Pixel Relationships, Generalized Structural Artifacts	Github

Continual & Incremental Learning

Methods that keep adapting detectors to evolving generators without catastrophic forgetting.

Title	Venue & Year	Modality	Highlights/Keywords	Code
Preserving Knowledge across Space and Time for Continual Video Deepfake Detection	ECCV 2026	`[V]`	Modality-Specific Frequency Distillation, Spatial/Temporal/Spatiotemporal Decomposition, Cross-Modality Decorrelation	GitHub
Automated In-the-Wild Data Collection for Continual AI Generated Image Detection	Arxiv 2026	`[I]`	Continual Learning, Continual Data Collection	GitHub
IncreFA: Breaking the Static Wall of Generative Model Attribution	Arxiv 2026	`[I]`	Incremental Learning, Generative Model Attribution	GitHub
SAIDO: Generalizable Detection of AI-Generated Images via Scene-Aware and Importance-Guided Dynamic Optimization in Continual Learning	CVPR 2026	`[I]`	Scene-aware optimization, Continual learning	GitHub
Generalizable and Adaptive Continual Learning Framework for AI-generated Image Detection	TMM 2026	`[I]`	Continual Learning, Kronecker-Factored Approximate Curvature	N/A

Papers related to AI-generated content safety (e.g., provenance/watermarking, misinformation verification) that do not fit the subcategories above.

Title	Venue & Year	Modality	Highlights/Keywords	Code
DF26: We Cannot Tell Fake From Real Anymore	Arxiv 2026	`[V]`	[Benchmark] 2,691 Public-Speaking Videos, 7 Modern T2V/I2V Models, Human & SOTA Detectors Near Chance, Distribution-Shift Robustness	N/A
APT: Anchor-aligned Perturbations for Tamper Localization in Fully Regenerated Images	ECCV 2026	`[I]`	[Proactive Forensics] Semi-Fragile Latent Perturbation, Fully Regenerated (Inpainting) Setting, Anchor-Direction Alignment	N/A
Training-Free Reconstruction-Based AI-Generated Image Detectors Are Inherently Vulnerable to Adversarial Examples	ECCV 2026 Workshop	`[I]`	[Robustness Analysis] Reconstruction-based Detector Attacks, Transferable Adversarial Examples, Real-world Degradations	N/A
Can We Defend Against AI-Generated Video Attacks on Real-World Crisis Events? A Systematic Evaluation of Detectors, Generators and Social Dissemination	Arxiv 2026	`[V]`	[Evaluation] RA-Bench, Crisis Event Videos, Detector Generalization, Social Dissemination	N/A
When Seeing Is Not Believing -- A Benchmark for Search-Grounded Video Misinformation Detection	Arxiv 2026	`[V]`	Search-Grounded Verification, EVID-Bench, Evidence-Dependent Manipulation	N/A
Robust ASIC-Based Image Authentication Using Reed-Solomon LSB Watermarking	Preprint 2026	`[I]`	ASIC PoW, Hardware-bound Provenance, Reed-Solomon Watermarking	GitHub

⬆ Back to Top

Competitions

Competition	Link	Year	Info
Robust AIGC Detection	NTIRE 2026 Robust AI-Generated Image Detection in the Wild	2026	No restrictions on training data. Evaluate ROC AUC metrics on robust samples.
Robust Deepfake Detection	NTIRE 2026 Robust Deepfake Detection Challenge	2026	No restrictions on training data.
The 6th Face Anti-spoofing Challenge	The 6th Face Anti-Spoofing: Unified Physical-Digital Attacks Detection@ICCV2025	2025	No external data or pre-trained models allowed. Limited to a single DL model with under 100G FLOPs.
Detect AI vs. Human-Generated Images	2025 Women in AI (WAI) Kaggle Challenge	2025	Paired dataset of authentic and AI-generated images
The 5th Face Anti-spoofing Challenge	5th Chalearn Face Anti-spoofing Workshop and Challenge@CVPR2024	2024	UniAttackData+ for unified physical and digital attack detection.

⬆ Back to Top

Practical Detection Tools

美亚鉴真 - 微信小程序搜索 美亚鉴真
SiliconSignature - GitHub - Hardware-bound image authentication using ASIC PoW nonces for unforgeable provenance certification
EyeSift - Website - Free online AI text/image/video/audio detector with detailed per-model benchmarks
Hive Moderation - Website
Tencent Zhuque AI Detection Assistant - Website
AI or Not - Website
Illuminarty - Website
Winston AI - Website
Is it AI? - Website
TruthScan - Website
中科睿鉴 (Zhongke Ruijian) - 微信小程序搜索 睿鉴AI

⬆ Back to Top

🏢 About Our Team

We are the Content Security Intelligence Team under Ant Group - Machine Intelligence. We are responsible for developing comprehensive content security and risk-mitigation capabilities for the Ant Group ecosystem, bridging the gap between rapidly evolving technologies and the urgent need for digital trust.

Why We Do It

In an era where synthetic media is increasingly sophisticated and pervasive, our research serves as a critical line of defense. By advancing AIGC detection technologies, we aim to:

Safeguard Digital Integrity: We provide essential defense mechanisms to protect the authenticity of visual content and combat the spread of misinformation in the digital space.
Empower Trust: Our solutions ensure the public can distinguish between genuine and synthetic media, fostering a more transparent and trustworthy digital ecosystem.
Industrial Application & Impact: We provide robust, scalable aigc detection solutions for Ant Group’s diverse content platforms, including Lingguang, Jingtan, and many others.

🤝 Collaborators

We are honored to collaborate with esteemed researchers and scholars in the field of AI and Computer Vision. We deeply value these academic partnerships that drive our innovation:

Prof. Jun Wan (万军) | CASIA & UCAS
- Research Interests: Biometrics, Face Anti-spoofing, Gesture Recognition, and Computer Vision.
- [Homepage]
Prof. Jianfu Zhang (张健夫) | Shanghai Jiao Tong University
- Research Interests: Computer Vision, Pattern Recognition, and Image/Video Analysis & Synthesis.
- [Homepage]
Prof. Zhuosheng Zhang (张倬胜) | Shanghai Jiao Tong University
- Research Interests: Natural Language Processing, Large Language Models, and Multi-modal Learning.
- [Homepage]

📝 Academic Publications

Veritas++: Value-aware On-Policy Distillation for Perception-Enhanced AIGI Detection | arXiv, 2026
- Highlights: Strengthened fine-grained visual perception for explainable AIGI detection through perception-oriented learning and value-aware on-policy distillation.
- [Paper] [Code]
VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning | ICML'26, 2026
- Highlights: Detected AI-generated videos using perception pretext reinforcement learning to capture temporal inconsistencies.
- [Paper] [Code]
Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated Images | CVPR'26, 2026
- Highlights: Improved detection accuracy through a two-stage approach of localizing suspicious regions followed by detailed examination.
- [Code]
GAMMA: Generalizable Alignment via Multi-task and Manipulation-Augmented Training for AI-Generated Image Detection | ICASSP'26, 2026
- Highlights: Enhanced generalization through multi-task learning and manipulation-augmented training strategies.
- [Paper]
FakeXplain: AI-Generated Image Detection via Human-Aligned Grounded Reasoning | ICLR'26, 2026
- Highlights: Detected AI-generated images through human-aligned grounded reasoning, providing interpretable visual evidence.
- [Paper] [Code]
Veritas: Generalizable deepfake detection via pattern-aware reasoning | ICLR'26 Oral, 2026
- Highlights: Achieved generalizable deepfake detection through pattern-aware reasoning, improving robustness across diverse manipulation types.
- [Paper] [Code]
Generalizable and Adaptive Continual Learning Framework for AI-generated Image Detection | IEEE TMM, 2025
- Highlights: Proposed a continual learning framework that adapts to new generative models while mitigating catastrophic forgetting.
- [Paper]
Towards explainable fake image detection with multi-modal large language models | ACM MM'25, 2025
- Highlights: Leveraged multi-modal large language models to provide human-interpretable explanations for fake image detection.
- [Paper]
WildFake: A Large-scale Challenging Dataset for AI-Generated Images Detection | AAAI'25 Oral, 2024
- Highlights: Introduced the largest and most comprehensive AIGC image dataset at the time, providing a challenging benchmark for detection models.
- [Paper]

🏆 Competition Achievements

1st Place Winner | NTIRE 2026 Robust AI-Generated Image Detection in the Wild Challenge
- Secured the top rank in ROC AUC for delivering superior performance in large-scale, real-world AI-generated image detection.
- [Challenge Website]
1st Place Winner | ICCV 2025 VQualA Challenge - Image Super-Resolution Generated Content Quality Assessment, 2025
- Achieved top performance in the VQualA 2025 challenge focused on assessing the quality of super-resolution generated content.
- [Paper 1] [Paper 2]
1st Place Winner | CVPR 2024 Face Anti-Spoofing Challenge, 2024
- Secured first place in the prestigious Face Anti-Spoofing Challenge at CVPR 2024, demonstrating state-of-the-art detection capabilities.
- [Challenge Website]

🛠️ Open-Source Resources

WildFake - A large and comprehensive AIGC image detection dataset.
- [ModelScope]
GenVideo - A large and comprehensive AIGC video detection dataset.
- [ModelScope]
HydraFake - A large-scale challenging dataset for AI-generated image detection.
- [ModelScope]
MintVid - A comprehensive video dataset for AIGC detection research.
- [ModelScope]

✉️ Contact Us

For questions or collaborations, please contact:

⬆ Back to Top

Star History

aigc-detection

ai-generated-detection

ai-generated-image-detection

awesome-list

multimodal-large-language-models

Contributors

26 commits

22 commits

3 commits

1 commits

Similar repos

MuskAI/Awesome-AIGC-Detection

HorizonTEL/AIGIBench

Purdue-M2/Detect-LAIM-generated-Multimedia-Survey

This repository contains a collection of resources and papers on Detecting Multimedia Generated by…

119

AI-Safeguard/Ivy-Fake

Dataset

IVY-FAKE: Unified Explainable Benchmark and Detector for AIGC Content

greatzh/Papers

Image Forgery Detection and Localization (and related) Papers List

571

HorizonTEL/AIGIBench

Dataset

Ziqiang Li1, Jiazhen Yan1, Ziwen He1, Kai Zeng2, Weiwei Jiang1, Lizhi Xiong1, Zhangjie Fu1‡

HorizonTEL/AIGIBench

Model

Ziqiang Li1, Jiazhen Yan1, Ziwen He1, Kai Zeng2, Weiwei Jiang1, Lizhi Xiong1, Zhangjie Fu1‡

hzlsaber/So-Fake

The offical repository of "So-Fake: Benchmarking and Explaining Social Media Image Forgery…

ant-research/Awesome-AIGC-Image-Video-Detection

A curated collection of the latest research and resources on AI-Generated Image and Video Detection.

459

52 commits

updated Sep 22, 2026

github.com/ant-research/Awesome-AIGC-Image-Video-Detection

See the code

README

Awesome AIGC Image/Video Detection

A curated collection of the latest research and resources on AI-Generated Image and Video Detection. This repository encompasses datasets, benchmarks, research papers, and practical detection tools.

🚀🚀🚀Contributions are welcome! If you find any missing papers, datasets, or tools, feel free to open an issue or submit a pull request.

Hot Events
Benchmarks & Datasets
Research Papers
- MLLM-Based
- Classification-Based
Competitions
Practical Detection Tools
About Our Team

🔥 Hot Events

Benchmarks & Datasets

Modality Legend: [I] Image | [V] Video | [M] Multi-modal

Annotation Type Legend: Au: Authenticity | Ex: Explainability | Lo: Localization

Benchmark	Paper	Venue & Year	Modality	Notes	Real Source	Fake Source/Generator	Annotation	Scale	Download
DF26	DF26: We Cannot Tell Fake From Real Anymore	Arxiv 2026	`[V]`	Modern-Generator AIGV Benchmark, Single-Person Public-Speaking Scenarios (Direct-to-Camera / Official Statement / Studio Interview), per-clip scene-prompt control (identity & scene fixed, generator varied), human + SOTA detector study near random chance, closed-source subset evaluation-only	OpenVid-1M, TalkingCelebs, MAVOS-DD	7 recent T2V/I2V models: Wan2.6, Veo 3.1, Grok Imagine 1.0, Kling 3.0 (closed); Wan2.2-A14B, HunyuanVideo 1.5, LTX 2.3 (open)	`Au`	2.7K (271 real + 2,420 fake, 1280×720)	DF26 (controlled access)
DailyBench	DailyBench: A Unified Benchmark for AI-Generated and Manipulated Images from Modern Generative Models	Arxiv 2026 (v3 2026-09-01)	`[I]`	Unified AIID testbed = FakeBench (full T2I synthesis) + ManipulationBench (object-level edits on real images); LAION-Aesthetics V2 pool filtered by aesthetic score ≥6 & shortest side ≥512; per-generator real/fake pairs; robustness study under recompression & pixel perturbation; ships FPD diagnostic baseline	LAION-Aesthetics V2	T2I: SD3.5-Large, FLUX.1, FLUX.2, Qwen-Image-2512, Z-Image, Nano Banana 2, GPT-Image 2; Edit: FLUX-Fill (random/object mask), FLUX.2-klein-9B, Qwen-Image-Edit-2511, Step1X-Edit-v1p2	`Au`	270K（v3：FakeBench ≈185K + ManipulationBench ≈75K）	DailyBench Project
AGIDefect-4K	AGIDefect-4K: A Richly Annotated Dataset for AI-Generated Image Defect Detection, Localization and Explanation	ACM MM 2026	`[I]`	Defect Detection, Localization & Explanation, 15 SOTA Generators, Quality Scoring		DALL-E 3, Midjourney, FLUX, Gemini, GPT-Image, Ideogram, Kling, Grok, etc.	`Au`, `Lo`, `Ex`	4K	AGIDefect-4K
RA-Bench	Can We Defend Against AI-Generated Video Attacks on Real-World Crisis Events? A Systematic Evaluation of Detectors, Generators and Social Dissemination	Arxiv 2026	`[V]`	Real-Crisis-Anchored, Source-Matched Evaluation, Human-Proof Subset, Propagation Robustness	Real crisis event footage (public media & source URLs)	4 open-source + 5 closed-source generators (incl. Wan2.2)	`Au`	17.9K	RA-Bench
RealHD	RealHD: A High-Quality Dataset for Robust Detection of State-of-the-Art AI-Generated Images	Arxiv 2026	`[I]`	Multi-category, SOTA Generators, 10K+ Prompts, Inpainting Masks		T2I, Inpainting, Refinement, Face Swapping	`Au`, `Lo`	730K+	RealHD
Treasure	Fleet: Few Shots Lead Effective AI-generated Image Detection	ICML 2026	`[I]`	64 Models, 20 Closed-source Commercial Engines, Few-shot Adaptation		Diverse architectures & 20 commercial engines	`Au`	360K	Treasure
LADBench	LADBench: A Benchmark for Logical Fault Detection in Images	ICDL 2026	`[I]`	Logical Anomaly Detection, VLM Evaluation, Common Sense Reasoning		Synthetic images with logical anomalies	`Au`	1K+	LADBench
EVID-Bench	When Seeing Is Not Believing -- A Benchmark for Search-Grounded Video Misinformation Detection	Arxiv 2026	`[V]`	Search-Grounded Verification, Evidence-Dependent Manipulation			`Au`		EVID-Bench
CoCoVideo	CoCoVideo: The High-Quality Commercial-Model-Based Contrastive Benchmark for AI-Generated Video Detection	Arxiv 2026	`[V]`	Commercial AIGC Models, Contrastive Benchmark		Commercial video generation models	`Au`		CoCoVideo
FraudBench	FraudBench: A Multimodal Benchmark for Detecting AI-Generated Fraudulent Refund Evidence	Arxiv 2026	`[M]`	Fraudulent Refund Detection (ECommerce)			`Au`		FraudBench
GPT-Image-2 Wild	GPT-Image-2 in the Wild: A Twitter Dataset of Self-Reported AI-Generated Images from the First Week of Deployment	Arxiv 2026	`[I]`	GPT-Image-2	Twitter (real images)	GPT-Image-2	`Au`	10K	GPT-Image-2 Wild
Artifact-Bench	Artifact-Bench: Evaluating MLLMs on Detecting and Assessing the Artifacts of AI-Generated Videos	Arxiv 2026	`[V]`	MLLM Evaluation, Video Artifacts			`Au`		Artifact-Bench
CommGen15	PGC: Peak-Guided Calibration for Generalizable AI-Generated Image Detection	ICML 2026	`[I]`	15 Commercial Generative Models		15 commercial generators	`Au`		CommGen15
AEGIS-Academic	AEGIS: A Holistic Benchmark for Evaluating Forensic Analysis of AI-Generated Academic Images	Arxiv 2026	`[I]`	Academic Image Forensics			`Au`		AEGIS
SciFigDetect	SciFigDetect: A Benchmark for AI-Generated Scientific Figure Detection	Arxiv 2026	`[I]`	Scientific Figure Detection		Nano Banana Pro, GPT-image-1.5	`Au`	150K	SciFigDetect
ActivityForensics	ActivityForensics: A Comprehensive Benchmark for Localizing Manipulated Activity in Videos	CVPR 2026	`[V]`	Action-level AIGC in videos			`Au`	6K	ActivityForensics
MintVid	VideoVeritas: AI-Generated Video Detection via Perception Pretext Reinforcement Learning	ICML 2026	`[V]`		OpenVid, VFHQ, HDTF, TikTok	Jimeng3.0-Pro, Seedance, Kling2.5-Turbo, Sora2, TikTok, Youtube, etc.	`Au`	4K	MintVid
AIGVDBench	Your One-Stop Solution for AI-Generated Video Detection	CVPR 2026	`[V]`		OpenVid-HD	31 generation models	`Au`	440k	AIGVDBench
HydraFake	Veritas: Generalizable Deepfake Detection via Pattern-Aware Reasoning	ICLR 2026(Oral)	`[I]`		FFHQ, VFHQ, CelebAHQ, FF++, etc.	GPT-4o, HailuoAI, ICLight, InfiniteYou, etc.	`Au`, `Ex`	100K	HydraFake
BR-Gen	Zooming In on Fakes: A Novel Dataset for Localized AI-Generated Image Detection with Forgery Amplification Approach	AAAI 2026	`[I]`				`Au`, `Lo`	150K	BR-Gen
RRDataset	Bridging the Gap Between Ideal and Real-world Evaluation: Benchmarking AI-Generated Image Detection in Challenging Scenarios	ICCV 2025	`[I]`	Real-World Robustness, Internet Transmission, Re-digitization			`Au`		DownLoad
HiResolution	No Pixel Left Behind: A Detail-Preserving Architecture for Robust High-Resolution AI-Generated Image Detection	ICLR 2026	`[I]`				`Au`	50K	HiRes-50K
AIGI-Now	AlignGemini: Generalizable AI-Generated Image Detection Through Task-Model Alignment	Arxiv 2026	`[I]`		COCO	Nano Banana, GPT-4o, Jimeng, Kling, Minimax, etc.	`Au`	18K	AIGI-Now
RealChain	Beyond Artifacts: Real-Centric Envelope Modeling for Reliable AI-Generated Image Detection	Arxiv 2026	`[I]`				`Au`	14K	RealChain
GenVidBench	GenVidBench: A 6-Million Benchmark for AI-Generated Video Detection	AAAI 2026	`[V]`				`Au`	6M	GenVidBench
Skyra	Skyra: AI-Generated Video Detection via Grounded Artifact Reasoning	CVPR 2026	`[V]`				`Au`, `Ex`, `Lo`	4K	ViF-CoT-4K
So-Fake-Set	So-Fake: Benchmarking and Explaining Social Media Image Forgery Detection	Arxiv 2025	`[I]`		F30k, WIDER, FFHQ, CelebA, OpenImages, COCO, OpenForensics	Qwen-image, GPT-4o, Nano Banana, Seedream3.0, Ideogram3.0, etc.	`Au`	2M+	So-Fake-Set So-Fake-OOD
GenBuster++	BusterX++: Towards Unified Cross-Modal AI-Generated Content Detection and Explanation with MLLM	Arxiv 2025	`[M]`				`Au`	4K	GenBuster++
GenBuster	BusterX: MLLM-Powered AI-Generated Video Forgery Detection and Explanation	Arxiv 2025	`[I]`				`Au`	200K	GenBuster-200K
AIGIBench	Is Artificial Intelligence Generated Image Detection a Solved Problem?	NeurIPS 2025	`[I]`		FFHQ, CelebA-HQ, Open Images V7	Common generators & SocialRF, CommunityAI	`Au`	200K	AIGIBench
Ivy-Fake	IVY-FAKE: A Unified Explainable Framework and Benchmark for Image and Video AIGC Detection	Arxiv 2025	`[M]`				`Au`, `Ex`	150K	Ivy-Fake
AEGIS	AEGIS: Authenticity Evaluation Benchmark for AI-Generated Video Sequences	ACM MM 2025	`[V]`		Vript (YouTube, TikTok), DVF, YouTube (self-collected)	Stable Video Diffusion, CogVideoX-5B, I2VGen-XL, Pika, KLing, Sora	`Au`, `Ex`	10K+	AEGIS
NeXT-IMDL	NeXT-IMDL: Build Benchmark for Next-Generation Image Manipulation Detection & Localization	Arxiv 2025	`[I]`		Flickr30k, COCO, OpenImages V7	SD2-Inpainting, SDXL-Inpainting, FLUX-Inpainting, etc.	`Au`, `Lo`	558K	NeXT-IMDL
ARForensics	D3QE: Learning Discrete Distribution Discrepancy-aware Quantization Error for Autoregressive-Generated Image Detection	ICCV 2025	`[I]`		ImageNet	Infinity, Janus_Pro, RAR, Switti, VAR, LlamaGen, Open_MAGVIT2	`Au`	300k	ARForensics
OpenSDI	OpenSDI: Spotting Diffusion-Generated Images in the Open World	CVPR 2025	`[I]`		Megalith-10M	SD1.5, SD2.1, SDXL, SD3, Flux.1	`Au`, `Lo`	300K	OpenSDI
Community Forensics	Community Forensics: Using Thousands of Generators to Train Fake Image Detectors	CVPR 2025	`[I]`		LAION, ImageNet, COCO, FFHQ, CelebA, MetFaces, AFHQ, etc.	4803 generators (Latent Diffusion, GAN, Autoregressive, Pixel Diffusion, Commercial)	`Au`	2.7M	Community Forensics
FakeClue	Spot the Fake: Large Multimodal Model-Based Synthetic Image Detection with Artifact Explanation	NeurIPS 2025	`[I]`				`Au`, `Ex`	100K	FakeClue
XAIGID-RewardBench	Explainable AI-Generated Image Detection RewardBench	NeurIPS 2025 Workshop	`[I]`		COCO-2017	Imagen 4, Flux.1 Dev, Bagel, etc.	`Au`, `Ex`	3K	XAIGID-RewardBench
RewardData	Learning Human-Perceived Fakeness in AI-Generated Videos via Multimodal LLMs	Arxiv 2025	`[V]`				`Au`, `Ex`	4.3K	RewardData
OpenFake	OPENFAKE: An Open Dataset and Platform Toward Real-World Deepfake Detection	Arxiv 2025	`[I]`		LAION-400M	SD 1.5/2.1/XL/3.5, Flux 1.0-dev/1.1-Pro/Schnell, Midjourney v6/v7, DALL·E 3, Imagen 3/4, GPT Image 1, Ideogram 3.0, Grok-2, HiDream-I1, Recraft v3, Chroma, and 10 community LoRA/finetune variants	`Au`	~4M	OPENFAKE
Video Reality Test	Video Reality Test: Can AI-Generated ASMR Videos fool VLMs and Humans?	Arxiv 2025	`[V]`		YouTube ASMR (social media)	Veo3.1-Fast, Sora2, Wan2.2-A14B, Wan2.2-5B, OpenSora-V2, HunyuanVideo, StepVideo	`Au`	149 real + dynamic fake	Video Reality Test
DDL	DDL: A Dataset for Interpretable Deepfake Detection and Localization in Real-World Scenarios	Arxiv 2025	`[M]`				`Au`	367K	DDL
DiffSeg30k	DiffSeg30k: A Multi-Turn Diffusion Editing Benchmark for Localized AIGC Detection	Arxiv 2025	`[I]`		COCO	SD2, SD3.5, SDXL, Flux.1, Glide, Kolors, HunyuanDiT1.1, Kandinsky 2.2	`Au`, `Lo`	30K	DiffSeg30k
FakeParts	FakeParts: a New Family of AI-Generated DeepFakes	Arxiv 2025	`[V]`				`Au`, `Lo`	81K	FakeParts
ForensicHub	ForensicHub: A Unified Benchmark & Codebase for All-Domain Fake Image Detection and Localization	NeurIPS 2025	`[I]`			ProGAN, StyleGAN, LDM, SDv1.4, SDv1.5, SDv2, SDXL, SD-ControlNet, MidJourney, ADM, GLIDE, VQDM, BigGAN	`Au`, `Lo`	23 datasets 42 models	ForensicHub
LOKI	LOKI: A Comprehensive Synthetic Data Detection Benchmark Using Large Multimodal Models	ICLR 2025	`[M]`			SORA, Keling, Open-Sora, FLUX, Midjourney, Stable Diffusion, Nerf-based, Gaussian-based, GPT-4o, Qwen-Max, Llama 3.1-405B, MusicGen, AudioLDM2...	`Au`, `Ex`	18K	LOKI
Chameleon	A Sanity Check for AI-Generated Image Detection	ICLR 2025	`[I]`		Unsplash	Midjourney, DALLE-3, Stable Diffusion (various LoRA fine-tuned)	`Au`	26K	Chameleon
WildFake	WildFake: A Large-scale Challenging Dataset for AI-Generated Images Detection	AAAI 2025	`[I]`				`Au`	3.7M	WildFake
WildRF	Real-Time Deepfake Detection in the Real-World	Arxiv 2024	`[I]`		Reddit, X (Twitter), Facebook (real images)	Reddit, X (Twitter), Facebook (social media deepfakes)	`Au`		WidlRF
AIGCDetectBenchmark	PatchCraft: Exploring Texture Patch for Efficient AI-generated Image Detection	Arxiv 2024	`[I]`				`Au`	100K	AIGCDetectionBenchMark
GenVideo	DeMamba: AI-Generated Video Detection on Million-Scale GenVideo Benchmark	Arxiv 2024	`[V]`				`Au`	2.3M	GenVideo
DRCT	Drct: Diffusion reconstruction contrastive training towards universal detection of diffusion generated images	ICML 2024	`[I]`		MSCOCO	LDM, SDv1.4, SDv1.5, SDv2, SDXL, SD-ControlNet	`Au`	2M	DRCT-2M
GenImage	GenImage: A Million-Scale Benchmark for Detecting AI-Generated Image	NeurIPS 2023	`[I]`		ImageNet, Wukong	MidJourney, SDv1.4, SDv1.5, ADM, GLIDE, VQDM, BigGAN	`Au`	2.7M	GenImage
DF40	DF40: Toward Next-Generation Deepfake Detection	NeurIPS 2024	`[I]` `[V]`				`Au`	0.1M+ videos, 1M+ images	DF40
Forensics-Bench	Forensics-Bench: A Comprehensive Forgery Detection Benchmark Suite for Large Vision Language Models	CVPR 2025	`[I]`, `[V]`, `[M]`		Various public datasets	GAN, Diffusion, VAE, RNN, Encoder-Decoder, Graphics-based	`Au`, `Lo`	63K	Forensics-Bench

⬆ Back to Top