video content moderators

Abstract: Vision-language models now grade robot videos, and text placed beside the frames can change their verdicts. On clips where the robot is losing progress, we add one log line saying the robot is making progress. Across Gemini, Qwen and GPT judges, this flips correct verdicts to progress 10 to 35 percentage points more often than a neutral line does. A one-line warning that the log may be wrong removes most of this effect. Selected fake "task complete" claims flip half of Qwen's correct verdicts, against a tenth for unrelated log lines. They have no detectable effect on Gemini.

Abstract: Self-blended images are widely used to train face-swap detectors, but primarily capture blending artifacts. We investigate whether adding illumination inconsistencies improves detection. Temporal Self-Blended Images (T-SBI) transfer lighting statistics between frames of the same video, with the mismatch controlled by luminance difference (ΔL). Using five training regimes and a three-seed comparison of high- and low-ΔL training, we find no evidence of illumination-specific improvements. AUC differences remain within seed variability across four datasets, and an analysis of 506,328 attribute-binned samples shows no preferential reduction in errors under harsh lighting. Instead, T-SBI shifts prediction scores, changing optimal thresholds by approximately 0.34 on FaceForensics++ and 0.30 on Celeb-DF, making comparisons at a fixed threshold misleading. However, T-SBI improves robustness to heavy JPEG compression on DFDC (AUC 0.780 versus 0.696), potentially reflecting greater reliance on low-frequency cues. These findings highlight the importance of evaluating training methods against their intended targets and accounting for threshold effects.

Abstract: Video deepfakes targeting a specific individual, the Person-of-Interest (POI), are the most harmful ones, and, since a public figure is abundantly recorded, a detector can be built from genuine footage of that individual. Such detectors commonly describe a subject through a 3D Morphable Model (3DMM) and adopt its coefficients as a whole, so which part of that description carries the signal has never been measured. We dissect it, holding the encoder, the training corpus and the enrollment protocol fixed and varying only what the encoder observes. The groups of coefficients prove largely redundant, since the shape block alone recovers almost all the accuracy of the full vector, and their temporal evolution contributes a real but bounded amount. We further show that the dense surface the same fit returns, which these detectors discard, carries identity information that the coefficients do not, and that it helps precisely where they are weakest. We assemble the best configuration into MOTIF, a visual-only detector trained on real videos only, with no manipulated video and no POI-specific data. It improves on both state-of-the-art POI detectors in every dataset and manipulation of our benchmark and at two quality levels. Our experimental code will be released at https://github.com/polimi-ispl/MOTIF.

Abstract: Large language models (LLMs) are increasingly used to assess long-form medical videos, but their effectiveness may depend on whether transcripts are provided in full or compressed through summarization, retrieval, or claim screening. This study examines how such transcript compression affects LLM-based veracity classification of Japanese medical YouTube videos. We compare four transcript input designs: full transcripts, LLM-generated summaries, RAPTOR-based retrievalaugmented generation (RAG), and Screening, which extracts candidate medical and health-related sentences. Using 74 long-form videos labeled as Real or Fake, we evaluate classification performance and analyze linguistic changes using J-LIWC, hedge expressions, and institutional or technical terms. The full-transcript Baseline achieved the best performance, whereas all compressed inputs increased false negatives, meaning that Fake videos were more likely to be misclassified as Real. Summary caused the largest performance drop, while Screening performed best among the compressed inputs but still omitted many medically relevant sentences. Linguistic analyses showed that these errors were not explained by a simple increase in certainty. Instead, Summary reduced affective, social, temporal, cognitive, and conversational cues, while Summary and RAG made institutional and technical terms more salient. These findings suggest that transcript compression can represent Fake videos as more coherent and authoritative inputs, thereby weakening cues needed for misinformation detection