Research map / Vision-language and multimodal / Multimodal LLMs
Trustworthy Image Authentication using Forensic Knowledge Graphs
Tai Nguyen, Matthew Stamm
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 self-supervised fingerprint learning to the backbone (ablation) | FKG · FKG-50K | detection AUC: 0.72 -> 0.94 | Improved |
| add self-supervised fingerprint learning to the backbone (ablation) | FKG · FKG-50K | localization F1: 0.44 -> 0.94 | Improved |
| add pairwise similarity constraint to contrastive loss (ablation) | FKG · FKG-50K | type accuracy: 0.63 -> 0.87 | Improved |
| add global self-attention to local graph attention in FRPN (ablation) | FKG · FKG-50K | type accuracy: 0.57 -> 0.87 | Improved |
| add local graph attention to full self-attention in FRPN (ablation) | FKG · FKG-50K | type accuracy: 0.55 -> 0.87 | Improved |
| add Hungarian matching loss to contrastive loss (ablation) | FKG · FKG-50K | localization F1: 0.72 -> 0.94 | Improved |
| add contrastive loss to Hungarian loss (ablation) | FKG · FKG-50K | detection AUC: 0.81 -> 0.94 | Improved |
| add transformer reasoner instead of linear classifiers (ablation) | FKG · FKG-50K | type accuracy: 0.69 -> 0.87 | Improved |
| add random in-context examples to the VLM prompt (ablation) | VLM justification · FKG-50K | completeness: 0.32 -> 0.51 | Improved |
| add chain-of-thought to random in-context examples (ablation) | VLM justification · FKG-50K | correctness: 0.63 -> 0.70 | Improved |
| use ICR protocol instead of random examples with CoT (ablation) | VLM justification · FKG-50K | completeness: 0.55 -> 0.85 | Improved |
| show forensic graph explanation alongside the verdict | FKG · human trust study (30 participants) | trust rating (1-5): 3.39 -> 4.53 | Improved |
Closest papers by meaning
- Forensics-Bench: A Comprehensive Forgery Detection Benchmark Suite for Large Vision Language Models (CVPR 2025 · similarity 0.86)
- Identity-Aware Vision-Language Model for Explainable Face Forgery Detection (AAAI 2026 · similarity 0.85)
- Community Forensics: Using Thousands of Generators to Train Fake Image Detectors (CVPR 2025 · similarity 0.85)
- Locate-Then-Examine: Grounded Region Reasoning Improves Detection of AI-Generated Images (CVPR 2026 · similarity 0.85)
- OmniVL-Guard: Towards Unified Vision-Language Forgery Detection and Grounding via Balanced RL (ICML 2026 · similarity 0.84)
- A Hitchhiker's Guide to Fine-Grained Face Forgery Detection Using Common Sense Reasoning (NeurIPS 2024 · similarity 0.84)
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.
