# Evidence Packing for Cross-Domain Image Deepfake Detection with LVLMs

arXiv: https://arxiv.org/abs/2603.17761 · HTML: https://mlsanity.com/paper/2603.17761/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 27 of its 37 references that are in the map.
- First posted on arXiv 18 Mar 2026.

## Builds on (most cited first, 27 of 27)

- [Qwen3 Technical Report](https://mlsanity.com/paper/2505.09388/) (2025)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [Gemma 3 Technical Report](https://mlsanity.com/paper/2503.19786/) (2025)
- [DynamicViT: Efficient Vision Transformers with Dynamic Token Sparsification](https://mlsanity.com/paper/2106.02034/) (2021)
- [Janus-Pro: Unified Multimodal Understanding and Generation with Data and Model Scaling](https://mlsanity.com/paper/2501.17811/) (2025)
- [mPLUG-Owl2: Revolutionizing Multi-modal Large Language Model with Modality Collaboration](https://mlsanity.com/paper/2311.04257/) (2023)
- [Janus: Decoupling Visual Encoding for Unified Multimodal Understanding and Generation](https://mlsanity.com/paper/2410.13848/) (2024)
- [Kimi-VL Technical Report](https://mlsanity.com/paper/2504.07491/) (2025)
- [AdaViT: Adaptive Tokens for Efficient Vision Transformer](https://mlsanity.com/paper/2112.07658/) (2021)
- [KonIQ-10k: An ecologically valid database for deep learning of blind image quality assessment](https://mlsanity.com/paper/1910.06180/) (2019)
- [PixArt-Sigma: Weak-to-Strong Training of Diffusion Transformer for 4K Text-to-Image Generation](https://mlsanity.com/paper/2403.04692/) (2024)
- [Infinity: Scaling Bitwise AutoRegressive Modeling for High-Resolution Image Synthesis](https://mlsanity.com/paper/2412.04431/) (2024)
- [Playground v2.5: Three Insights towards Enhancing Aesthetic Quality in Text-to-Image Generation](https://mlsanity.com/paper/2402.17245/) (2024)
- [MiniCPM-V 4.5: Cooking Efficient MLLMs via Architecture, Data, and Training Recipe](https://mlsanity.com/paper/2509.18154/) (2025)
- [Deepfake Generation and Detection: A Benchmark and Survey](https://mlsanity.com/paper/2403.17881/) (2024)
- [Common Sense Reasoning for Deepfake Detection](https://mlsanity.com/paper/2402.00126/) (2024)
- [Delving into Sequential Patches for Deepfake Detection](https://mlsanity.com/paper/2207.02803/) (2022)
- [Frequency Decoupling for Motion Magnification via Multi-Level Isomorphic Architecture](https://mlsanity.com/paper/2403.07347/) (2024)
- [Kandinsky 3: Text-to-Image Synthesis for Multifunctional Generative Framework](https://mlsanity.com/paper/2410.21061/) (2024)
- [Deepfake Detection that Generalizes Across Benchmarks](https://mlsanity.com/paper/2508.06248/) (2025)
- [MMAD: Multi-label Micro-Action Detection in Videos](https://mlsanity.com/paper/2407.05311/) (2024)
- [Bridging Different Language Models and Generative Vision Models for Text-to-Image Generation](https://mlsanity.com/paper/2403.07860/) (2024)
- [LookupViT: Compressing visual information to a limited number of tokens](https://mlsanity.com/paper/2407.12753/) (2024)
- [FreqDebias: Towards Generalizable Deepfake Detection via Consistency-Driven Frequency Debiasing](https://mlsanity.com/paper/2509.22412/) (2025)
- [DFBench: Benchmarking Deepfake Image Detection Capability of Large Multimodal Models](https://mlsanity.com/paper/2506.03007/) (2025)
- [PRPO: Paragraph-level Policy Optimization for Vision-Language Deepfake Detection](https://mlsanity.com/paper/2509.26272/) (2025)
- [Visual Language Models as Zero-Shot Deepfake Detectors](https://mlsanity.com/paper/2507.22469/) (2025)

## Abstract (arXiv)

Image Deepfake Detection (IDD) separates manipulated images from authentic ones by spotting artifacts of synthesis or tampering. Although large vision-language models (LVLMs) offer strong image understanding, adapting them to IDD often demands costly fine-tuning and generalizes poorly to diverse, evolving manipulations. We propose the Semantic Consistent Evidence Pack (SCEP), a training-free LVLM framework that replaces whole-image inference with evidence-driven reasoning. SCEP mines a compact set of suspicious patch tokens that best reveal manipulation cues. It uses the vision encoder's CLS token as a global reference, clusters patch features into coherent groups, and scores patches with a fused metric combining CLS-guided semantic mismatch with frequency-and noise-based anomalies. To cover dispersed traces and avoid redundancy, SCEP samples a few high-confidence patches per cluster and applies grid-based NMS, producing an evidence pack that conditions a frozen LVLM for prediction. Experiments on diverse benchmarks show SCEP outperforms strong baselines without LVLM fine-tuning.