Search

Recent advancements in text-to-image diffusion models have brought them to the public spotlight, becoming widely accessible and embraced by everyday users. However, these models have been shown to generate harmful content such as not-safe-for-work (NSFW) images. While approaches have been proposed to erase such abstract concepts from the models, jail-breaking techniques have succeeded in bypassing such safety measures. In this paper, we propose TraSCE, an approach to guide the diffusion trajectory away from generating harmful content. Our approach is based on negative prompting, but as we show in this paper, conventional negative prompting is not a complete solution and can easily be bypassed in some corner cases. To address this issue, we first propose a modification of conventional negative prompting. Furthermore, we introduce a localized loss-based guidance that enhances the modified negative prompting technique by steering the diffusion trajectory. We demonstrate that our proposed method achieves state-of-the-art results on various benchmarks in removing harmful content including ones proposed by red teams; and erasing artistic styles and objects. Our proposed approach does not require any training, weight modifications, or training data (both image or prompt), making it easier for model owners to erase new concepts.

Diffusion models are prone to exactly reproduce images from the training data. This exact reproduction of the training data is concerning as it can lead to copyright infringement and/or leakage of privacy-sensitive information. In this paper, we present a novel way to understand the memorization phenomenon, and propose a simple yet effective approach to mitigate it. We argue that memorization occurs because of an attraction basin in the denoising process which steers the diffusion trajectory towards a memorized image. However, this can be mitigated by guiding the diffusion trajectory away from the attraction basin by not applying classifier-free guidance until an ideal transition point occurs from which classifier-free guidance is applied. This leads to the generation of non-memorized images that are high in image quality and well-aligned with the conditioning mechanism. To further improve on this, we present a new guidance technique, \emph{opposite guidance}, that escapes the attraction basin sooner in the denoising process. We demonstrate the existence of attraction basins in various scenarios in which memorization occurs, and we show that our proposed approach successfully mitigates memorization.

Postdoctoral training is a career stage often described as a demanding and anxiety-laden time when many promising PhDs see their academic dreams slip away due to circumstances beyond their control. We use a unique data set of academic publishing and careers to chart the more or less successful postdoctoral paths. We build a measure of academic success on the citation patterns two to five years into a faculty career. Then, we monitor how students' postdoc positions -- in terms of relocation, change of topic, and early well-cited papers -- relate to their early-career success. One key finding is that the postdoc period seems more important than the doctoral training to achieve this form of success. This is especially interesting in light of the many studies of academic faculty hiring that link Ph.D. granting institutions and hires, omitting the postdoc stage. Another group of findings can be summarized as a Goldilocks principle: it seems beneficial to change one's direction, but not too much.

This article explores the theoretical underpinnings of -- and practical results for -- topological and geometric features of data collected by synthetic aperture sonar systems. We prove a strong theoretical guarantee about the structure of the space of echos (the signature space) that is relevant for classification, and that this structure is measurable in laboratory conditions. The guarantee makes minimal assumptions about the sonar platform's trajectory, and establishes that the signature space divides neatly into topological features based upon the number of prominent echos and their placement, and geometric features that capture their corresponding sonar cross section., Comment: 34 pages, 16 figures

Social robot navigation (SRN) is a relevant problem that involves navigating a pedestrian-rich environment in a socially acceptable manner. It is an essential part of making social robots effective in pedestrian-rich settings. The use cases of such robots could vary from companion robots to warehouse robots to autonomous wheelchairs. In recent years, deep reinforcement learning has been increasingly used in research on social robot navigation. Our work introduces CAMRL (Context-Aware Mamba-based Reinforcement Learning). Mamba is a new deep learning-based State Space Model (SSM) that has achieved results comparable to transformers in sequencing tasks. CAMRL uses Mamba to determine the robot's next action, which maximizes the value of the next state predicted by the neural network, enabling the robot to navigate effectively based on the rewards assigned. We evaluate CAMRL alongside existing solutions (CADRL, LSTM-RL, SARL) using a rigorous testing dataset which involves a variety of densities and environment behaviors based on ORCA and SFM, thus, demonstrating that CAMRL achieves higher success rates, minimizes collisions, and maintains safer distances from pedestrians. This work introduces a new SRN planner, showcasing the potential for deep-state space models for robot navigation.

Open Domain Question Answering (ODQA) within natural language processing involves building systems that answer factual questions using large-scale knowledge corpora. Recent advances stem from the confluence of several factors, such as large-scale training datasets, deep learning techniques, and the rise of large language models. High-quality datasets are used to train models on realistic scenarios and enable the evaluation of the system on potentially unseen data. Standardized metrics facilitate comparisons between different ODQA systems, allowing researchers to objectively track advancements in the field. Our study presents a thorough examination of the current landscape of ODQA benchmarking by reviewing 52 datasets and 20 evaluation techniques across textual and multimodal modalities. We introduce a novel taxonomy for ODQA datasets that incorporates both the modality and difficulty of the question types. Additionally, we present a structured organization of ODQA evaluation metrics along with a critical analysis of their inherent trade-offs. Our study aims to empower researchers by providing a framework for the robust evaluation of modern question-answering systems. We conclude by identifying the current challenges and outlining promising avenues for future research and development., Comment: 22 pages, 13 tables, 7 figures

We demonstrate absorber-free passive and hybrid mode-locking at sub-GHz repetition rates using a hybrid integrated extended cavity diode laser around 1550 nm. The laser is based on InP as gain medium and a long Si$_3$N$_4$ feedback circuit, with three highly frequency selective microring resonators. The feedback resonators not only increases the cavity length up to 0.6 m to achieve sub-GHz repetition rates but also serve as a dispersive narrowband mirror for sharp spectral filtering, which enables Fourier domain mode-locking. We observe passive mode-locking with repetition rates below 500 MHz, with $\approx$ 15 comb lines at around 0.2 mW total power. To stabilize the repetition rate, hybrid mode-locking is demonstrated by weak RF modulation of the diode current at frequencies around 500 MHz. The RF injection reduces the Lorentzian linewidth component from 8.9 kHz to a detection limited value around 300 mHz. To measure the locking range of the repetition rate, the injected RF frequency is tuned with regard to the passive mode-locking frequency and the injected RF power is varied. The locking range increases approximately as a square-root function of the injected RF power. At 1 mW injection a wide locking range of about 80 MHz is obtained. We observe the laser maintaining stable mode-locking also when the DC diode pump current is increased from 40 mA to 190 mA, provided that the cavity length is maintained constant with thermo-refractive tuning., Comment: This manuscript is under review for publication with optics express

Through a large-scale study over diverse face images, we show that facial attribute editing using modern generative AI models can severely degrade automated face recognition systems. This degradation persists even with identity-preserving generative models. To mitigate this issue, we propose two novel techniques for local and global attribute editing. We empirically ablate twenty-six facial semantic, demographic and expression-based attributes that have been edited using state-of-the-art generative models, and evaluate them using ArcFace and AdaFace matchers on CelebA, CelebAMaskHQ and LFW datasets. Finally, we use LLaVA, an emerging visual question-answering framework for attribute prediction to validate our editing techniques. Our methods outperform the current state-of-the-art at facial editing (BLIP, InstantID) while improving identity retention by a significant extent., Comment: Under review

The rise of AI voice-cloning technology, particularly audio Real-time Deepfakes (RTDFs), has intensified social engineering attacks by enabling real-time voice impersonation that bypasses conventional enrollment-based authentication. To address this, we propose PITCH, a robust challenge-response method to detect and tag interactive deepfake audio calls. We developed a comprehensive taxonomy of audio challenges based on the human auditory system, linguistics, and environmental factors, yielding 20 prospective challenges. These were tested against leading voice-cloning systems using a novel dataset comprising 18,600 original and 1.6 million deepfake samples from 100 users. PITCH's prospective challenges enhanced machine detection capabilities to 88.7% AUROC score on the full unbalanced dataset, enabling us to shortlist 10 functional challenges that balance security and usability. For human evaluation and subsequent analyses, we filtered a challenging, balanced subset. On this subset, human evaluators independently scored 72.6% accuracy, while machines achieved 87.7%. Acknowledging that call environments require higher human control, we aided call receivers in making decisions with them using machines. Our solution uses an early warning system to tag suspicious incoming calls as "Deepfake-likely." Contrary to prior findings, we discovered that integrating human intuition with machine precision offers complementary advantages. Our solution gave users maximum control and boosted detection accuracy to 84.5%. Evidenced by this jump in accuracy, PITCH demonstrated the potential for AI-assisted pre-screening in call verification processes, offering an adaptable and usable approach to combat real-time voice-cloning attacks. Code to reproduce and access data at \url{https://github.com/mittalgovind/PITCH-Deepfakes}.

In this commentary, we discuss the evolving nature of search engines, as they begin to generate, index, and distribute content created by generative artificial intelligence (GenAI). Our discussion highlights challenges in the early stages of GenAI integration, particularly around factual inconsistencies and biases. We discuss how output from GenAI carries an unwarranted sense of credibility, while decreasing transparency and sourcing ability. Furthermore, search engines are already answering queries with error-laden, generated content, further blurring the provenance of information and impacting the integrity of the information ecosystem. We argue how all these factors could reduce the reliability of search engines. Finally, we summarize some of the active research directions and open questions., Comment: 9 pages, 5 figures