Introduces the Impostor benchmark dataset for localizing AIGC image manipulations via agent curation and the PANet model that uses phase and semantic consistency for better detection.
hub
A sanity check for ai-generated image detection
28 Pith papers cite this work. Polarity classification is still indexing.
28
Pith papers citing it
abstract
With the rapid development of generative models, discerning AI-generated content has evoked increasing attention from both industry and academia. In this paper, we conduct a sanity check on "whether the task of AI-generated image detection has been solved". To start with, we present Chameleon dataset, consisting AIgenerated images that are genuinely challenging for human perception. To quantify the generalization of existing methods, we evaluate 9 off-the-shelf AI-generated image detectors on Chameleon dataset. Upon analysis, almost all models classify AI-generated images as real ones. Later, we propose AIDE (AI-generated Image DEtector with Hybrid Features), which leverages multiple experts to simultaneously extract visual artifacts and noise patterns. Specifically, to capture the high-level semantics, we utilize CLIP to compute the visual embedding. This effectively enables the model to discern AI-generated images based on semantics or contextual information; Secondly, we select the highest frequency patches and the lowest frequency patches in the image, and compute the low-level patchwise features, aiming to detect AI-generated images by low-level artifacts, for example, noise pattern, anti-aliasing, etc. While evaluating on existing benchmarks, for example, AIGCDetectBenchmark and GenImage, AIDE achieves +3.5% and +4.6% improvements to state-of-the-art methods, and on our proposed challenging Chameleon benchmarks, it also achieves the promising results, despite this problem for detecting AI-generated images is far from being solved.
hub tools
citation-role summary
baseline 2
background 1
citation-polarity summary
representative citing papers
SynCred-Bench shows that 15 MLLMs reach only 10.5% TPR, open-source detectors under 5%, commercial APIs 57.6%, and humans 63% TPR at 5% FPR when identifying AI-generated images with synthetic credibility.
ReAlign distills LLM-generated reasoning texts into a lightweight AIGI forgery detector via contrastive image-text alignment to improve generalization on complex forgeries.
FakeReasoning is an MLLM-based framework for unified forgery detection and reasoning on AI-generated images, supported by the new MMFR-Dataset of 120K images and 378K annotations across 10 generators.
A 36-model cross-paradigm benchmark on a hard 100-image corpus shows commercial APIs lead on MCC, open-source detectors trail on average, and a subset of strong rankers are miscalibrated at their default threshold.
SalArt-VQA benchmark shows that high image-level artifact detection accuracy in VLMs does not imply correct localization, grounding, or evidence-supported defect descriptions.
Color transformations expose statistical discrepancies in synthetic images, supporting a classifier with 93.27% average accuracy and robustness to post-processing.
Social gaze consistency between interacting people is proposed as a new semantic cue orthogonal to low-level artifacts for detecting AI-generated images, with reported accuracy gains on vision and vision-language models.
HydraPrompt uses an Asymmetric Prompt Adapter with fixed real prompts and adaptive fake prompts plus a Conditional Supervised Contrastive loss to achieve SOTA synthetic image detection on benchmarks.
SEF introduces GAN upsampling for diverse artifacts and expert fusion to reduce domain interference, yielding stronger generalization on 13 benchmarks for AI-generated image detection.
ODP-Net uses instance-aware orthogonal decomposition, perturbation-based purification, and manifold alignment to separate universal forgery traces, generator fingerprints, and semantics, achieving SOTA on unseen architectures like Stable Diffusion 3.
A detector that trains a separate LoRA adapter per generator and mixes them with a learned router reports the best mean accuracy on AIGIBench and Chameleon, with weaker per-dataset results on several subsets.
Intermediate layer embedding sensitivity to perturbations distinguishes AI-generated images from real ones, yielding higher AUROC on GenImage and Forensics Small benchmarks than prior methods.
Frozen features from vision foundation models enable a linear probe to outperform specialized AIGI detectors by over 30% on in-the-wild data due to emergent forgery knowledge from pre-training.
GAPL learns a compact set of canonical forgery prototypes and applies two-stage LoRA training to build a low-variance feature space that improves generalization across GAN and diffusion generators.
PiN-CLIP jointly trains a noise generator and detector under a variational positive-incentive principle to inject feature-space noise that suppresses shortcut directions and improves out-of-distribution accuracy by 5.4 points on images from 42 generative models.
The ITW-SM dataset and targeted optimization of detector design choices yield a 26.87% average AUC improvement for state-of-the-art AI-generated image detectors under real-world social media conditions.
HFI detects LDM-generated images without training data by quantifying aliasing in autoencoder outputs and supports model-specific implicit watermarking.
Frozen multimodal encoders enable robust AI-generated image detection via linear classification on a 10K-image curated training set that improves generalization over larger datasets.
MDMF detects AI-generated images by learning patch-level forensic signatures and quantifying their distributional discrepancies with MMD, yielding larger separation than global methods when micro-defects are present.
FakeVLM-R1 combines GRPO reinforcement learning with critical-thinking CoT and a physics-annotated FakeClue++ dataset to reach claimed SOTA synthetic image detection while reducing over-rejection of real images.
SPECTRA-Net fuses multi-view tensor representations from vision foundation models, spectral analysis, local anomaly detection, and statistical descriptors to achieve state-of-the-art cross-domain AI-generated image detection with explainable artifact localization.
A 3B-parameter vision-language model trained on continuously curated social media data detects AI-generated content with state-of-the-art accuracy on benchmarks and shows positive engagement effects in production deployment.
I2P adaptively selects the most discriminative layers from visual foundation models for synthetic image detection and constrains task updates to low-sensitivity parameter subspaces to improve specificity without harming generalization.
citing papers explorer
-
Impostor: An Agent-Curated Benchmark for Realistic AIGC Manipulation Localization
Introduces the Impostor benchmark dataset for localizing AIGC image manipulations via agent curation and the PANet model that uses phase and semantic consistency for better detection.
-
SynCred-Bench: Benchmarking Synthetic Credibility in AI-Generated Visual Misinformation
SynCred-Bench shows that 15 MLLMs reach only 10.5% TPR, open-source detectors under 5%, commercial APIs 57.6%, and humans 63% TPR at 5% FPR when identifying AI-generated images with synthetic credibility.
-
ReAlign: Generalizable Image Forgery Detection via Reasoning-Aligned Representation
ReAlign distills LLM-generated reasoning texts into a lightweight AIGI forgery detector via contrastive image-text alignment to improve generalization on complex forgeries.
-
Toward Generalizable Forgery Detection and Reasoning
FakeReasoning is an MLLM-based framework for unified forgery detection and reasoning on AI-generated images, supported by the new MMFR-Dataset of 120K images and 378K annotations across 10 generators.
-
VendorBench-100: A Unified Cross-Paradigm Benchmark for Deepfake Image Detection
A 36-model cross-paradigm benchmark on a hard 100-image corpus shows commercial APIs lead on MCC, open-source detectors trail on average, and a subset of strong rankers are miscalibrated at their default threshold.
-
SalArt-VQA: Diagnosing Whether VLMs Understand Salient Artifacts in Generated Images
SalArt-VQA benchmark shows that high image-level artifact detection accuracy in VLMs does not imply correct localization, grounding, or evidence-supported defect descriptions.
-
Chroma Clues: Leveraging Color Statistics to Detect Synthetic Images
Color transformations expose statistical discrepancies in synthetic images, supporting a classifier with 93.27% average accuracy and robustness to post-processing.
-
When Eyes Betray AI: Social Gaze Consistency as a Semantic Cue for AI-Generated Image Detection
Social gaze consistency between interacting people is proposed as a new semantic cue orthogonal to low-level artifacts for detecting AI-generated images, with reported accuracy gains on vision and vision-language models.
-
HydraPrompt: An Adaptive and Asymmetric Framework of Vision-Language Models for Synthetic Image Detection
HydraPrompt uses an Asymmetric Prompt Adapter with fixed real prompts and adaptive fake prompts plus a Conditional Supervised Contrastive loss to achieve SOTA synthetic image detection on benchmarks.
-
Reduce the Artifacts Bias for More Generalizable AI-Generated Image Detection
SEF introduces GAN upsampling for diverse artifacts and expert fusion to reduce domain interference, yielding stronger generalization on 13 benchmarks for AI-generated image detection.
-
Decoupling Semantics and Fingerprints: A Universal Representation for AI-Generated Image Detection
ODP-Net uses instance-aware orthogonal decomposition, perturbation-based purification, and manifold alignment to separate universal forgery traces, generator fingerprints, and semantics, achieving SOTA on unseen architectures like Stable Diffusion 3.
-
LEGO: LoRA-Enabled Generator-Oriented Framework for Synthetic Image Detection
A detector that trains a separate LoRA adapter per generator and mixes them with a learned router reports the best mean accuracy on AIGIBench and Chameleon, with weaker per-dataset results on several subsets.
-
Intermediate Representations are Strong AI-Generated Image Detectors
Intermediate layer embedding sensitivity to perturbations distinguishes AI-generated images from real ones, yielding higher AUROC on GenImage and Forensics Small benchmarks than prior methods.
-
Simplicity Prevails: The Emergence of Generalizable AIGI Detection in Visual Foundation Models
Frozen features from vision foundation models enable a linear probe to outperform specialized AIGI detectors by over 30% on in-the-wild data due to emergent forgery knowledge from pre-training.
-
Scaling Up AI-Generated Image Detection with Generator-Aware Prototypes
GAPL learns a compact set of canonical forgery prototypes and applies two-stage LoRA training to build a low-variance feature space that improves generalization across GAN and diffusion generators.
-
How Noise Benefits AI-generated Image Detection
PiN-CLIP jointly trains a noise generator and detector under a variational positive-incentive principle to inject feature-space noise that suppresses shortcut directions and improves out-of-distribution accuracy by 5.4 points on images from 42 generative models.
-
Navigating the Challenges of AI-Generated Image Detection in the Wild: What Truly Matters?
The ITW-SM dataset and targeted optimization of detector design choices yield a 26.87% average AUC improvement for state-of-the-art AI-generated image detectors under real-world social media conditions.
-
HFI: A unified framework for training-free detection and implicit watermarking of latent diffusion model generated images
HFI detects LDM-generated images without training data by quantifying aliasing in autoencoder outputs and supports model-specific implicit watermarking.
-
SSAFE: Simple and Strong AI-Generated Image Detection via Frozen Vision Encoders
Frozen multimodal encoders enable robust AI-generated image detection via linear classification on a 10K-image curated training set that improves generalization over larger datasets.
-
Micro-Defects Expose Macro-Fakes: Detecting AI-Generated Images via Local Distributional Shifts
MDMF detects AI-generated images by learning patch-level forensic signatures and quantifying their distributional discrepancies with MMD, yielding larger separation than global methods when micro-defects are present.
-
FakeVLM-R1: Internalizing Physical Laws via CoT for Synthetic Image Detection
FakeVLM-R1 combines GRPO reinforcement learning with critical-thinking CoT and a physics-annotated FakeClue++ dataset to reach claimed SOTA synthetic image detection while reducing over-rejection of real images.
-
SPECTRA-Net: Scalable Pipeline for Explainable Cross-domain Tensor Representations for AI-generated Images Detection
SPECTRA-Net fuses multi-view tensor representations from vision foundation models, spectral analysis, local anomaly detection, and statistical descriptors to achieve state-of-the-art cross-domain AI-generated image detection with explainable artifact localization.
-
Detecting AI-Generated Content on Social Media with Multi-modal Language Models
A 3B-parameter vision-language model trained on continuously curated social media data detects AI-generated content with state-of-the-art accuracy on benchmarks and shows positive engagement effects in production deployment.
-
Adaptive Forensic Feature Refinement via Intrinsic Importance Perception
I2P adaptively selects the most discriminative layers from visual foundation models for synthetic image detection and constrains task updates to low-sensitivity parameter subspaces to improve specificity without harming generalization.
-
Boosting Robust AIGI Detection with LoRA-based Pairwise Training
LoRA-based pairwise training with distortion and size simulations boosts robust AIGI detection under severe distortions, placing third in the NTIRE challenge.
-
NTIRE 2026 Challenge on Robust AI-Generated Image Detection in the Wild
The NTIRE 2026 challenge provides a dataset of over 294,000 real and AI-generated images with 36 transformations to benchmark robust detection models.
-
Deepfakes: we need to re-think the concept of "real" images
This position paper contends that the concept of 'real' images must be rethought because most modern photographs are computationally generated, undermining current deepfake detection methods.
- Dissect and Prune: Enhancing Robustness in AI-Generated Image Detection