Team MSU GenText-Forensics Challenge 2026 Technical Report

Team MSU placed third in the ACM MM 2026 GenText-Forensics challenge with a decomposed chain-of-thought pipeline that pairs a document tampering detector with two Qwen3-VL-32B vision-language models, each LoRA-adapted to a distinct sub-task. The detector produces tampering probability maps converted into numbered candidate regions; a Filterer model validates them and assigns a preliminary forgery type, while a Semantic Detective model merges and re-grounds surviving regions, hunts semantic anomalies invisible to pixel-level detectors, and writes the final forensic report. Both models were trained by distilling chain-of-thought traces from a privileged Qwen3-VL-235B teacher with access to ground-truth masks and reports, and the team reports ablations over detector thresholds, prompt designs, and pipeline decompositions.

arXiv:2609.38391v1 Announce Type: new Abstract: Document text forgery has evolved beyond simple pixel-level manipulation: modern attacks alter not only the appearance of a document but also its meaning, and increasingly target the OCR & LLM pipelines that consume such documents. The ACM MM 2026 GenText-Forensics challenge therefore requires systems that not only decide whether a multilingual text image is forged, but also localize the point of manipulation, identify the attack type, and produce a human-readable forensic report with supporting evidence. We present our solution, a decomposed chain-of-thought CoT pipeline that combines a document tampering detector DTD with two Qwen3-VL-32B vision-language models, each LoRA-adapted to a distinct sub-task. DTD produces tampering probability maps that are converted into numbered candidate regions; a first model the Filterer validates these regions and assigns a preliminary forgery type, while a second model the Semantic Detective merges and re-grounds the surviving regions, searches for purely semantic anomalies that are invisible to pixel-level detectors, and writes the final report. Both models are trained by distilling chain-of-thought traces from a privileged Qwen3-VL-235B teacher that has access to ground-truth masks and reports. Our approach secured third place in the ACM MM 2026 GenText-Forensics challenge. We describe the data preparation, test-time augmentation, region rendering, distillation protocol, and training configuration in detail, and report ablations over detector thresholds, prompt designs, and pipeline decompositions.