Learning Human-Perceived Fakeness in AI-Generated Videos via Multimodal LLMs

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Fu, Xingyu, Liu, Siyi, Xu, Yinuo, Lu, Pan, Hu, Guangqiuse, Yang, Tianbo, Anantasagar, Taran, Shen, Christopher, Mao, Yikai, Liu, Yuanzhe, Shah, Keyush, Lee, Chung Un, Choi, Yejin, Zou, James, Roth, Dan, Callison-Burch, Chris
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911186906251264
author Fu, Xingyu
Liu, Siyi
Xu, Yinuo
Lu, Pan
Hu, Guangqiuse
Yang, Tianbo
Anantasagar, Taran
Shen, Christopher
Mao, Yikai
Liu, Yuanzhe
Shah, Keyush
Lee, Chung Un
Choi, Yejin
Zou, James
Roth, Dan
Callison-Burch, Chris
author_facet Fu, Xingyu
Liu, Siyi
Xu, Yinuo
Lu, Pan
Hu, Guangqiuse
Yang, Tianbo
Anantasagar, Taran
Shen, Christopher
Mao, Yikai
Liu, Yuanzhe
Shah, Keyush
Lee, Chung Un
Choi, Yejin
Zou, James
Roth, Dan
Callison-Burch, Chris
contents Can humans identify AI-generated (fake) videos and provide grounded reasons? While video generation models have advanced rapidly, a critical dimension -- whether humans can detect deepfake traces within a generated video, i.e., spatiotemporal grounded visual artifacts that reveal a video as machine generated -- has been largely overlooked. We introduce DeeptraceReward, the first fine-grained, spatially- and temporally- aware benchmark that annotates human-perceived fake traces for video generation reward. The dataset comprises 4.3K detailed annotations across 3.3K high-quality generated videos. Each annotation provides a natural-language explanation, pinpoints a bounding-box region containing the perceived trace, and marks precise onset and offset timestamps. We consolidate these annotations into 9 major categories of deepfake traces that lead humans to identify a video as AI-generated, and train multimodal language models (LMs) as reward models to mimic human judgments and localizations. On DeeptraceReward, our 7B reward model outperforms GPT-5 by 34.7% on average across fake clue identification, grounding, and explanation. Interestingly, we observe a consistent difficulty gradient: binary fake v.s. real classification is substantially easier than fine-grained deepfake trace detection; within the latter, performance degrades from natural language explanations (easiest), to spatial grounding, to temporal labeling (hardest). By foregrounding human-perceived deepfake traces, DeeptraceReward provides a rigorous testbed and training signal for socially aware and trustworthy video generation.
format Preprint
id arxiv_https___arxiv_org_abs_2509_22646
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Learning Human-Perceived Fakeness in AI-Generated Videos via Multimodal LLMs
Fu, Xingyu
Liu, Siyi
Xu, Yinuo
Lu, Pan
Hu, Guangqiuse
Yang, Tianbo
Anantasagar, Taran
Shen, Christopher
Mao, Yikai
Liu, Yuanzhe
Shah, Keyush
Lee, Chung Un
Choi, Yejin
Zou, James
Roth, Dan
Callison-Burch, Chris
Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
Can humans identify AI-generated (fake) videos and provide grounded reasons? While video generation models have advanced rapidly, a critical dimension -- whether humans can detect deepfake traces within a generated video, i.e., spatiotemporal grounded visual artifacts that reveal a video as machine generated -- has been largely overlooked. We introduce DeeptraceReward, the first fine-grained, spatially- and temporally- aware benchmark that annotates human-perceived fake traces for video generation reward. The dataset comprises 4.3K detailed annotations across 3.3K high-quality generated videos. Each annotation provides a natural-language explanation, pinpoints a bounding-box region containing the perceived trace, and marks precise onset and offset timestamps. We consolidate these annotations into 9 major categories of deepfake traces that lead humans to identify a video as AI-generated, and train multimodal language models (LMs) as reward models to mimic human judgments and localizations. On DeeptraceReward, our 7B reward model outperforms GPT-5 by 34.7% on average across fake clue identification, grounding, and explanation. Interestingly, we observe a consistent difficulty gradient: binary fake v.s. real classification is substantially easier than fine-grained deepfake trace detection; within the latter, performance degrades from natural language explanations (easiest), to spatial grounding, to temporal labeling (hardest). By foregrounding human-perceived deepfake traces, DeeptraceReward provides a rigorous testbed and training signal for socially aware and trustworthy video generation.
title Learning Human-Perceived Fakeness in AI-Generated Videos via Multimodal LLMs
topic Computer Vision and Pattern Recognition
Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2509.22646