EFR trains a vision-language model to bind each forensic conclusion to specific image and text locations, improving detection and grounding on DGM4 while making explanations spatially verifiable.
Title resolution pending
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2026 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
Evidence-Grounded Forensic Reasoning for Detecting and Grounding Multi-Modal Media Manipulation
EFR trains a vision-language model to bind each forensic conclusion to specific image and text locations, improving detection and grounding on DGM4 while making explanations spatially verifiable.