Yufei Zhu

Aparna Ananthasubramaniam, Yufei 'Louise' Zhu, David Jurgens et al.

The diffusion of culture online is theorized to be influenced by many interacting social factors (e.g., network and identity). However, most existing computational cascade models consider just a single factor (e.g., network or identity). This work offers a new framework for teasing apart the mechanisms underlying hashtag cascades. We curate a new dataset of 1,337 hashtags representing cultural innovation online, develop a 10-factor evaluation framework for comparing empirical and simulated cascades, and show that a combined network+identity model better simulates hashtag cascades than network- or identity-only counterfactuals. We also explore heterogeneity in performance: While a combined network+identity model best predicts the popularity of cascades, a network-only model best predicts cascade growth and an identity-only model best predicts adopter composition. The network+identity model has the highest comparative advantage among hashtags used for expressing racial or regional identity and talking about sports or news. In fact, we are able to predict what combination of network and/or identity best models each hashtag and use this to further improve performance. Our results show the utility of models incorporating the interactions of network, identity, and other social factors in the diffusion of hashtags in social media.

Kuofeng Gao, Yufei Zhu, Yiming Li et al.

Text-to-image (T2I) diffusion models have rapidly advanced, enabling high-quality image generation conditioned on textual prompts. However, the growing trend of fine-tuning pre-trained models for personalization raises serious concerns about unauthorized dataset usage. To combat this, dataset ownership verification (DOV) has emerged as a solution, embedding watermarks into the fine-tuning datasets using backdoor techniques. These watermarks remain inactive under benign samples but produce owner-specified outputs when triggered. Despite the promise of DOV for T2I diffusion models, its robustness against copyright evasion attacks (CEA) remains unexplored. In this paper, we explore how attackers can bypass these mechanisms through CEA, allowing models to circumvent watermarks even when trained on watermarked datasets. We propose the first copyright evasion attack (i.e., CEAT2I) specifically designed to undermine DOV in T2I diffusion models. Concretely, our CEAT2I comprises three stages: watermarked sample detection, trigger identification, and efficient watermark mitigation. A key insight driving our approach is that T2I models exhibit faster convergence on watermarked samples during the fine-tuning, evident through intermediate feature deviation. Leveraging this, CEAT2I can reliably detect the watermarked samples. Then, we iteratively ablate tokens from the prompts of detected watermarked samples and monitor shifts in intermediate features to pinpoint the exact trigger tokens. Finally, we adopt a closed-form concept erasure method to remove the injected watermark. Extensive experiments show that our CEAT2I effectively evades DOV mechanisms while preserving model performance.