AniPrO: Interpretable Anime Image Provenance Detection via Multi-Dimensional Semantic Reasoning

Yan Liu; Baoxiang Huang; Zi'an Wang; Wenbo Xie

cs.CVSep 20, 2026

AniPrO: Interpretable Anime Image Provenance Detection via Multi-Dimensional Semantic Reasoning

Authors: Yan Liu, Baoxiang Huang, Zi'an Wang, Wenbo Xie

Organizations: School of Computer Science and Technology, Tongji University, Shanghai, China · College of Electronic and Information Engineering, Tongji University, Shanghai, China

Abstract

As generative AI becomes increasingly used in anime-style image creation, distinguishing human-drawn, AI-inpainted, and text-to-image images is important for copyright attribution, visual provenance, and content governance. Existing AI-generated image detectors mainly target real-world photographs and often overlook anime-specific cues such as flat coloring, exaggerated structures, and artistic line control. To address this gap, we propose AniPrO, a multi-dimensional description-enhanced framework for interpretable anime image provenance. Built upon AnimeDL-2M, AniPrO contains 15,000 balanced samples from a 35,000-image candidate pool, covering Real, Inpainting, and Text2Image categories with structured five-dimensional descriptions. We further introduce AniPrO-SFD-Bench and AniPrO-MFR-Bench to evaluate provenance detection from statistical feature discrimination and multimodal fusion reasoning perspectives. Experiments show that structured semantic guidance reveals systematic AI-generation biases, such as the gap between global visual plausibility and local detail coherence, and improves the detection of challenging inpainting samples. The dataset and code will be released at: https://github.com/YAN-LIU05/AniPrO.

Figures & tables

Figure 1: Overview of the proposed multi-dimensional description-enhanced framework. The pipeline consists of three main stages: (1) Dataset Construction of the AniPrO dataset; (2) Multi-dimensional Annotation, which extracts five-dimensional structured scores and texts using multimodal LLMs; and (3) A dual-track evaluation system utilizing Statistical Metrics and LLM-as-a-Judge for final AI vs. Real image discrimination.

Figure 2: Representative Real, Inpainting, and Text2Image samples from AniPrO with structured descriptions.

Features	Real Ret.	Inpainting Int.	T2I Int.	Overall Acc.
The Bucket Effect	$\mathbf{0.883}$	0.321	0.561	0.588
Analysis of Variance	0.862	0.300	0.492	0.551
Skewness Analysis	0.644	0.390	0.485	0.506
Gap Analysis	0.768	$\mathbf{0.495}$	$\mathbf{0.762}$	$\mathbf{0.675}$

Table 1: Univariate discriminatory performance based on derived statistical features using scores extracted from Gemini-

2.5

-Pro as an example.

Scoring Models	Real Ret.	Inpainting Int.	T2I Int.	Overall Acc.
GPT- $5.2$	$0.578\pm 0.018$	$0.465\pm 0.011$	$0.706\pm 0.015$	$0.588\pm 0.011$
Gemini- $2.5$ -Pro	$\mathbf{0.841}\pm\mathbf{0.014}$	$0.428\pm 0.021$	$0.838\pm 0.014$	$\mathbf{0.702}\pm\mathbf{0.007}$
Doubao- $1.5$ -Vision-Pro	$0.252\pm 0.009$	$0.754\pm 0.014$	$\mathbf{0.927}\pm\mathbf{0.009}$	$0.644\pm 0.003$
Qwen3-VL-235B	$0.572\pm 0.014$	$0.557\pm 0.017$	$0.788\pm 0.015$	$0.639\pm 0.007$
Qwen3-VL-30B	$0.612\pm 0.021$	$0.534\pm 0.030$	$0.783\pm 0.032$	$0.643\pm 0.015$
Llama- $4$ -Maverick	$0.459\pm 0.021$	$0.599\pm 0.026$	$0.764\pm 0.012$	$0.607\pm 0.009$

Table 2: Results of

5

-fold cross-validation based on Random Forest.

Description Extraction Models	Real Ret.	Inpainting Int.	T2I Int.	Overall Acc.
GPT- $5.2$	0.485	0.626	0.835	0.649
Gemini- $2.5$ -Pro	$\mathbf{0.584}$	0.657	0.967	$\mathbf{0.738}$
Doubao- $1.5$ -Vision-Pro	0.481	0.642	0.920	0.684
Qwen3-VL-235B	0.489	0.613	0.902	0.678
Qwen3-VL-30B	0.410	0.696	0.929	0.665
Llama- $4$ -Maverick	0.479	0.663	0.934	0.695

Table 3: Results of AniPrO-MFR-Bench with GLM-

4.5

V as the fixed judge.

Judge Models	Experimental Setup	Real Ret.	Inpainting Int.	T2I Int.	Overall Acc.
Gemini- $2.5$ -Pro	Image-Only	$\mathbf{0.986}$	0.181	0.968	0.807
Gemini- $2.5$ -Pro	Ours	0.938	$\mathbf{0.501}$	$\mathbf{0.970}$	$\mathbf{0.825}$
GLM- $4.5$ V	Image-Only	$\mathbf{0.817}$	0.296	0.697	0.585
GLM- $4.5$ V	Description-Only	0.630	0.522	0.838	0.663
GLM- $4.5$ V	Ours	0.584	$\mathbf{0.657}$	$\mathbf{0.967}$	$\mathbf{0.738}$

Table 4: Results of multimodal input ablation experiments. All descriptive texts are extracted uniformly by Gemini-

2.5

-Pro.

Setting	Real Ret.	Inpainting Int.	T2I Int.	Overall Acc.	$\Delta$ Acc.
Full 5D	0.584	0.657	$\mathbf{0.967}$	$\mathbf{0.738}$	–
w/o 3D Structures and Spatial Relationships	0.588	0.594	0.871	0.693	$-0.045$
w/o Drawing Logic and Manual Control	0.433	$\mathbf{0.704}$	0.894	0.678	$-0.060$
w/o Lighting and Textures	0.651	0.604	0.944	0.712	$-0.026$
w/o Fine Details	0.580	0.578	0.897	0.673	$-0.065$
w/o Contour and Line Quality	$\mathbf{0.653}$	0.556	0.804	0.704	$-0.034$

Table 5: Leave-one-dimension-out ablation results.

Method	Real Ret.	Inpainting Int.	T2I Int.	Overall Acc.
CNN baseline	0.522	0.656	0.982	0.720
ViT-B/16	0.835	0.211	0.551	0.532
Human raters	0.619	0.479	0.833	0.644

Table 6: Supervised visual classifiers and human rater baselines. Human results are averaged over five anime enthusiasts on a randomly sampled 10% subset.

AI-Generated Image Detection

AI-generated image detection identifies synthetic or AI-edited images and develops methods that generalize across generators, domains, and image transformations.

See all 194 papers

SPARED: Reasoning-Based AI-Generated Image Detection via Adversarially Edited Data
Aug 13, 2026Yicheng Bao, Xiahui Guo, Xuhong Wang +1Adversarial Training Adversarial Examples
1East China Normal University · 2Shanghai Artificial Intelligence Laboratory
Grounding and Explaining Visual Evidence for AI-Generated Image Detection in Human-Centric Scenes
Aug 3, 2026Kun Guo, Yuzhou Yang, Haoyue Wang +3Visual Grounding AI-Generated Image Detection
College of Computer Science and Artificial Intelligence, Fudan University · NVIDIA
Where Detectors Fail: Probing Generative Space for Generalizable AI-Generated Image Detection
May 24, 2026Zijie Cao, Weijie Tu, Yao Xiao +3Synthetic Data Augmentation AI-Generated Image Detection
School of Computer Science and Engineering, Sun Yat-sen University, Guangzhou, China · Australian National University, Canberra, Australia · Tsinghua Shenzhen International Graduate School, Tsinghua University, Shenzhen, China +1

Aug 13, 2026J/K move · Enter open · S save

SPARED: Reasoning-Based AI-Generated Image Detection via Adversarially Edited Data

Yicheng Bao, Xiahui Guo, Xuhong Wang, Xin Tan

1East China Normal University · 2Shanghai Artificial Intelligence Laboratory

Adversarial Training Adversarial Examples AI-Generated Image Detection Adversarial Attacks on AI-Generated Image Detectors

Open paper AskPDF arXiv

Detecting AI-generated images is only half the task: a deployed detector must also justify its verdict, yet existing detectors inherit three failure modes from their training data: real and fake images collected from different sources invite provenance shortcuts, supervised explanation corpora teach templated rationales, and a static forgery corpus leaves the decision boundary standing still while generators keep moving. We introduce \methodname{}, an adversarial reinforcement learning framework that pits two heterogeneous models against each other. A diffusion image editor learns to edit real photographs into fake counterparts of those same photographs that fool the current detector, while a reasoning MLLM learns to expose them with a verdict grounded in free-form reasoning. Both rewards are shortcut-proof by design: the attacker is credited only when its edit is faithfully executed, and the defender only when its verdict is correct. As the two models alternate, each round's attacker regenerates a harder training pool aimed at the current detector's blind spots, so the detector must generalize rather than memorize any fixed artifact distribution. Although the explanation is never rewarded, its quality rises round over round as a side effect of accuracy-only training. A detector trained within this loop improves monotonically across rounds on each of three external benchmarks.

Abstract

AI-Generated Image Detection

Explore similar work

SPARED: Reasoning-Based AI-Generated Image Detection via Adversarially Edited Data

Grounding and Explaining Visual Evidence for AI-Generated Image Detection in Human-Centric Scenes

Where Detectors Fail: Probing Generative Space for Generalizable AI-Generated Image Detection