Kazem Faghih, Wenxiao Wang, Yize Cheng, Siddhant Bharti, Gaurang Sriramanan, Sriram Balasubramanian, Parsa Hosseini, Soheil Feizi. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

The advent of text-to-image generative models has led artists to worry that their individual styles may be copied, creating a pressing need to reconsider the lack of protection for artistic styles under copyright law. This requires answering challenging questions, like what defines style and what constitutes style infringment. In this work, we build on prior legal scholarship to develop an automatic and interpretable framework to \emph{quantitatively} assess style infringement. Our methods hinge on a simple logical argument: if an artist's works can consistently be recognized as their own, then they have a unique style. Based on this argument, we introduce ArtSavant, a practical (i.e., efficient and easy to understand) tool to (i) determine the unique style of an artist by comparing it to a reference corpus of works from hundreds of artists, and (ii) recognize if the identified style reappears in generated images. We then apply ArtSavant in an empirical study to quantify the prevalence of artistic style copying across 3 popular text-to-image generative models, finding that under simple prompting, $20\\%$ of $372$ prolific artists studied appear to have their styles be at risk of copying by today's generative models. Our findings show that prior legal arguments can be operationalized in quantitative ways, towards more nuanced examination of the issue of artistic style infringements.

Recent work has explored how individual components of the CLIP-ViT model contribute to the final representation by leveraging the shared image-text representation space of CLIP. These components, such as attention heads and MLPs, have been shown to capture distinct image features like shape, color or texture. However, understanding the role of these components in arbitrary vision transformers (ViTs) is challenging. To this end, we introduce a general framework which can identify the roles of various components in ViTs beyond CLIP. Specifically, we (a) automate the decomposition of the final representation into contributions from different model components, and (b) linearly map these contributions to CLIP space to interpret them via text. Additionally, we introduce a novel scoring function to rank components by their importance with respect to specific features. Applying our framework to various ViT variants (e.g. DeiT, DINO, DINOv2, Swin, MaxViT), we gain insights into the roles of different components concerning particular image features. These insights facilitate applications such as image retrieval using text descriptions or reference images, visualizing token importance heatmaps, and mitigating spurious correlations. We release our [code](https://github.com/SriramB-98/vit-decompose) to reproduce the experiments in the paper.

Simulating physical network paths (e.g., Internet) is a cornerstone research problem in the emerging sub-field of AI-for-networking. We seek a model that generates end-to-end packet delay values in response to the time-varying load offered by a sender, which is typically a function of the previously output delays. The problem setting is unique, and renders the state-of-the-art text and time-series generative models inapplicable or ineffective. We formulate an ML problem at the intersection of dynamical systems, sequential decision making, and time-series modeling. We propose a novel grey-box approach to network simulation that embeds the semantics of physical network path in a new RNN-style model called Recurrent Buffering Unit, providing the interpretability of standard network simulator tools, the power of neural models, the efficiency of SGD-based techniques for learning, and yielding promising results on synthetic and real-world network traces.

The ability to remove features from the input of machine learning models is very important to understand and interpret model predictions. However, this is non-trivial for vision models since masking out parts of the input image typically causes large distribution shifts. This is because the baseline color used for masking (typically grey or black) is out of distribution. Furthermore, the shape of the mask itself can contain unwanted signals which can be used by the model for its predictions. Recently, there has been some progress in mitigating this issue (called missingness bias) in image masking for vision transformers. In this work, we propose a new masking method for CNNs we call layer masking in which the missingness bias caused by masking is reduced to a large extent. Intuitively, layer masking applies a mask to intermediate activation maps so that the model only processes the unmasked input. We show that our method (i) is able to eliminate or minimize the influence of the mask shape or color on the output of the model, and (ii) is much better than replacing the masked region by black or grey for input perturbation based interpretability techniques like LIME. Thus, layer masking is much less affected by missingness bias than other masking strategies. We also demonstrate how the shape of the mask may leak information about the class, thus affecting estimates of model reliance on class-relevant features derived from input masking. Furthermore, we discuss the role of data augmentation techniques for tackling this problem, and argue that they are not sufficient for preventing model reliance on mask shape.

Despite the remarkable success of deep neural networks in a myriad of settings, several works have demonstrated their overwhelming sensitivity to near-imperceptible perturbations, known as adversarial attacks. On the other hand, prior works have also observed that deep networks can be under-sensitive, wherein large-magnitude perturbations in input space do not induce appreciable changes to network activations. In this work, we study in detail the phenomenon of under-sensitivity in vision models such as CNNs and Transformers, and present techniques to study the geometry and extent of “equi-confidence” level sets of such networks. We propose a Level Set Traversal algorithm that iteratively explores regions of high confidence with respect to the input space using orthogonal components of the local gradients. Given a source image, we use this algorithm to identify inputs that lie in the same equi-confidence level set as the source image despite being perceptually similar to arbitrary images from other classes. We further observe that the source image is linearly connected by a high-confidence path to these inputs, uncovering a star-like structure for level sets of deep networks. Furthermore, we attempt to identify and estimate the extent of these connected higher-dimensional regions over which the model maintains a high degree of confidence.

Process-induced variations and sub-threshold leakage in bulk-Si technology limit the scaling of SRAM into sub-32 nm nodes. New device architectures are being considered to improve$V_{T}$control and reduce short channel effects. Among the likely candidates, FinFETs are the most attractive option because of their good scalability and possibilities for further SRAM performance and yield enhancement through independent gating. The enhancements to read/write margins and yield are investigated in detail for two cell designs employing independently gated FinFETs. It is shown that FinFET-based 6-T SRAM cells designed with pass-gate feedback (PGFB) achieve significant improvements in the cell read stability without area penalty. The write-ability of the cell can be improved through the use of pull-up write gating (PUWG) with a separate write word line (WWL). The benefits of these two approaches are complementary and additive, allowing for simultaneous read and write yield enhancements when the PGFB and PUWG designs are used in combination.

The requirement of low power and wide bandwidth operation in indoor environment implies that a multiband OFDM modem has to work both at low SNR and in presence of large delay spreads, posing severe challenges in designing acquisition functions. In this paper, we have described techniques of estimating and compensating carrier and clock frequency offset for multiband OFDM based UWB systems in presence of OFDM symbol to symbol band hopping , that yield reliable result at SNR of around 0 db and in presence of delay spread of up to 60 nsec. System simulation results have been shown to illustrate the robustness of our methods, and comparison have been made to theoretical lower bounds achievable.

The focus for ultra-wideband (UWB) technology has recently shifted to applications involving short-range, high-speed wireless communications like IEEE 802.15.3a for personal area network (PAN). FCC defines UWB signals as one that occupies more than 500 MHz bandwidth in the 3.1 to 10.6 GHz band and meets the output power limit of -41.3 dBm/MHz. This paper addresses the problem of equalizer design for multi-band OFDM based UWB systems. The novelty of the proposed solution is that it is based on joint MLSE equalization & diversity-combining for both QPSK & DCM modes and it performs reliably at low SNRs. The solution is amenable for VLSI implementation. Maximal-ratio-combining (MRC) based alternatives are also analyzed in details and it is shown that an MRC receiver for DCM is very sensitive to timing errors, which reduces its usefulness. Finally we develop alternatives to DCM that are robust to timing errors.

Intrinsic variations and challenging leakage control in today's bulk-Si MOSFETs limit the scaling of SRAM. Design tradeoffs in six-transistor (6-T) and four-transistor (4-T) SRAM cells are presented in this work. It is found that 6-T and 4-T FinFET-based SRAM cells designed with built-in feedback achieve significant improvements in the cell static noise margin (SNM) without area penalty. Up to 2x improvement in SNM can be achieved in 6-T FinFET-based SRAM cells. A 4-T FinFET-based SRAM cell with built-in feedback can achieve sub-100pA per-cell standby current and offer the similar improvements in SNM as the 6-T cell with feedback, making them attractive for low-power, low-voltage applications