Search

Ethical intervention prompting has emerged as a tool to counter demographic biases of text-to-image generative AI models. Existing solutions either require to retrain the model or struggle to generate images that reflect desired distributions on gender and race. We propose an inference-time process called DebiasPI for Debiasing-by-Prompt-Iteration that provides prompt intervention by enabling the user to control the distributions of individuals' demographic attributes in image generation. DebiasPI keeps track of which attributes have been generated either by probing the internal state of the model or by using external attribute classifiers. Its control loop guides the text-to-image model to select not yet sufficiently represented attributes, With DebiasPI, we were able to create images with equal representations of race and gender that visualize challenging concepts of news headlines. We also experimented with the attributes age, body type, profession, and skin tone, and measured how attributes change when our intervention prompt targets the distribution of an unrelated attribute type. We found, for example, if the text-to-image model is asked to balance racial representation, gender representation improves but the skin tone becomes less diverse. Attempts to cover a wide range of skin colors with various intervention prompts showed that the model struggles to generate the palest skin tones. We conducted various ablation studies, in which we removed DebiasPI's attribute control, that reveal the model's propensity to generate young, male characters. It sometimes visualized career success by generating two-panel images with a pre-success dark-skinned person becoming light-skinned with success, or switching gender from pre-success female to post-success male, thus further motivating ethical intervention prompting with DebiasPI., Comment: This work was presented at The European Conference on Computer Vision (ECCV) 2024 Workshop "Fairness and ethics towards transparent AI: facing the chalLEnge through model Debiasing" (FAILED), Milano, Italy, on September 29, 2024, https://failed-workshop-eccv-2024.github.io

Geocoding involves automatic extraction of location coordinates of incidents reported in news articles, and can be used for epidemic intelligence or disaster management. This paper introduces Retrieval-Augmented Coordinate Capture Of Online News articles (RACCOON), an open-source geocoding approach that extracts geolocations from news articles. RACCOON uses a retrieval-augmented generation (RAG) approach where candidate locations and associated information are retrieved in the form of context from a location database, and a prompt containing the retrieved context, location mentions and news articles is fed to an LLM to generate the location coordinates. Our evaluation on three datasets, two underlying LLMs, three baselines and several ablation tests based on the components of RACCOON demonstrate the utility of RACCOON. To the best of our knowledge, RACCOON is the first RAG-based approach for geocoding using pre-trained LLMs., Comment: Accepted at WWW 2025 as a short paper. 4 pages with references

Uncertainty calibration, the process of aligning confidence with accuracy, is a hallmark of human intelligence. However, most machine learning models struggle to achieve this alignment, particularly when the training dataset is small relative to the network's capacity. Here, we demonstrate that uncertainty calibration can be effectively achieved through a pretraining method inspired by developmental neuroscience. Specifically, training with random noise before data training allows neural networks to calibrate their uncertainty, ensuring that confidence levels are aligned with actual accuracy. We show that randomly initialized, untrained networks tend to exhibit erroneously high confidence, but pretraining with random noise effectively calibrates these networks, bringing their confidence down to chance levels across input spaces. As a result, networks pretrained with random noise exhibit optimal calibration, with confidence closely aligned with accuracy throughout subsequent data training. These pre-calibrated networks also perform better at identifying "unknown data" by exhibiting lower confidence for out-of-distribution samples. Our findings provide a fundamental solution for uncertainty calibration in both in-distribution and out-of-distribution contexts.

Visual-Language Models (VLMs) have become a powerful tool for bridging the gap between visual and linguistic understanding. However, the conventional learning approaches for VLMs often suffer from limitations, such as the high resource requirements of collecting and training image-text paired data. Recent research has suggested that language understanding plays a crucial role in the performance of VLMs, potentially indicating that text-only training could be a viable approach. In this work, we investigate the feasibility of enhancing fine-grained visual understanding in VLMs through text-only training. Inspired by how humans develop visual concept understanding, where rich textual descriptions can guide visual recognition, we hypothesize that VLMs can also benefit from leveraging text-based representations to improve their visual recognition abilities. We conduct comprehensive experiments on two distinct domains: fine-grained species classification and cultural visual understanding tasks. Our findings demonstrate that text-only training can be comparable to conventional image-text training while significantly reducing computational costs. This suggests a more efficient and cost-effective pathway for advancing VLM capabilities, particularly valuable in resource-constrained environments., Comment: AAAI25 workshop accepted

Diffusion-based Generative AI gains significant attention for its superior performance over other generative techniques like Generative Adversarial Networks and Variational Autoencoders. While it has achieved notable advancements in fields such as computer vision and natural language processing, their application in speech generation remains under-explored. Mainstream Text-to-Speech systems primarily map outputs to Mel-Spectrograms in the spectral space, leading to high computational loads due to the sparsity of MelSpecs. To address these limitations, we propose LatentSpeech, a novel TTS generation approach utilizing latent diffusion models. By using latent embeddings as the intermediate representation, LatentSpeech reduces the target dimension to 5% of what is required for MelSpecs, simplifying the processing for the TTS encoder and vocoder and enabling efficient high-quality speech generation. This study marks the first integration of latent diffusion models in TTS, enhancing the accuracy and naturalness of generated speech. Experimental results on benchmark datasets demonstrate that LatentSpeech achieves a 25% improvement in Word Error Rate and a 24% improvement in Mel Cepstral Distortion compared to existing models, with further improvements rising to 49.5% and 26%, respectively, with additional training data. These findings highlight the potential of LatentSpeech to advance the state-of-the-art in TTS technology

Recent advancements in text-to-speech (TTS) systems, such as FastSpeech and StyleSpeech, have significantly improved speech generation quality. However, these models often rely on duration generated by external tools like the Montreal Forced Aligner, which can be time-consuming and lack flexibility. The importance of accurate duration is often underestimated, despite their crucial role in achieving natural prosody and intelligibility. To address these limitations, we propose a novel Aligner-Guided Training Paradigm that prioritizes accurate duration labelling by training an aligner before the TTS model. This approach reduces dependence on external tools and enhances alignment accuracy. We further explore the impact of different acoustic features, including Mel-Spectrograms, MFCCs, and latent features, on TTS model performance. Our experimental results show that aligner-guided duration labelling can achieve up to a 16\% improvement in word error rate and significantly enhance phoneme and tone alignment. These findings highlight the effectiveness of our approach in optimizing TTS systems for more natural and intelligible speech generation.

The Large Vision Language Model (VLM) has recently addressed remarkable progress in bridging two fundamental modalities. VLM, trained by a sufficiently large dataset, exhibits a comprehensive understanding of both visual and linguistic to perform diverse tasks. To distill this knowledge accurately, in this paper, we introduce a novel approach that explicitly utilizes VLM as an objective function form for the Human-Object Interaction (HOI) detection task (\textbf{VLM-HOI}). Specifically, we propose a method that quantifies the similarity of the predicted HOI triplet using the Image-Text matching technique. We represent HOI triplets linguistically to fully utilize the language comprehension of VLMs, which are more suitable than CLIP models due to their localization and object-centric nature. This matching score is used as an objective for contrastive optimization. To our knowledge, this is the first utilization of VLM language abilities for HOI detection. Experiments demonstrate the effectiveness of our method, achieving state-of-the-art HOI detection accuracy on benchmarks. We believe integrating VLMs into HOI detection represents important progress towards more advanced and interpretable analysis of human-object interactions., Comment: 18 pages

Recent advancements in reinforcement learning (RL) for analog circuit optimization have demonstrated significant potential for improving sample efficiency and generalization across diverse circuit topologies and target specifications. However, there are challenges such as high computational overhead, the need for bespoke models for each circuit. To address them, we propose M3, a novel Model-based RL (MBRL) method employing the Mamba architecture and effective scheduling. The Mamba architecture, known as a strong alternative to the transformer architecture, enables multi-circuit optimization with distinct parameters and target specifications. The effective scheduling strategy enhances sample efficiency by adjusting crucial MBRL training parameters. To the best of our knowledge, M3 is the first method for multi-circuit optimization by leveraging both the Mamba architecture and a MBRL with effective scheduling. As a result, it significantly improves sample efficiency compared to existing RL methods.

Lyrics generation presents unique challenges, particularly in achieving precise syllable control while adhering to song form structures such as verses and choruses. Conventional line-by-line approaches often lead to unnatural phrasing, underscoring the need for more granular syllable management. We propose a framework for lyrics generation that enables multi-level syllable control at the word, phrase, line, and paragraph levels, aware of song form. Our approach generates complete lyrics conditioned on input text and song form, ensuring alignment with specified syllable constraints. Generated lyrics samples are available at: https://tinyurl.com/lyrics9999

Emerging Large Language Model (LLM) applications require long input prompts to perform complex downstream tasks like document analysis and code generation. For these long context length applications, the length of the input prompt poses a significant challenge in terms of inference efficiency since the inference costs increase linearly with sequence length. However, for many of these applications, much of the context in the prompt is fixed across different user inputs, thereby providing the opportunity to perform offline optimizations to process user inputs quickly, as they are received. In this work, we propose Squeezed Attention as a mechanism to accelerate LLM applications where a large portion of the input prompt is fixed. We first leverage K-means clustering offline to group the keys for the fixed context based on semantic similarity and represent each cluster with a single centroid value. During inference, we compare query tokens from the user input with the centroids to predict which of the keys from the fixed context are semantically relevant and need to be loaded during inference. We then compute exact attention using only these important keys from the fixed context, thereby reducing bandwidth and computational costs. We also extend our method to use a hierarchical centroid lookup to identify important keys, which can reduce the complexity of attention from linear to logarithmic with respect to the context length. We implement optimized Triton kernels for centroid comparison and sparse FlashAttention with important keys, achieving more than 4x speedups during both the prefill and generation phases for long-context inference. Furthermore, we have extensively evaluated our method on various long-context benchmarks including LongBench, where it achieves a 3x reduction in KV cache budget without accuracy loss and up to an 8x reduction with <0.5 point accuracy gap for various models.

Drone-captured images present significant challenges in object detection due to varying shooting conditions, which can alter object appearance and shape. Factors such as drone altitude, angle, and weather cause these variations, influencing the performance of object detection algorithms. To tackle these challenges, we introduce an innovative vision-language approach using learnable prompts. This shift from conventional manual prompts aims to reduce domain-specific knowledge interference, ultimately improving object detection capabilities. Furthermore, we streamline the training process with a one-step approach, updating the learnable prompt concurrently with model training, enhancing efficiency without compromising performance. Our study contributes to domain-generalized object detection by leveraging learnable prompts and optimizing training processes. This enhances model robustness and adaptability across diverse environments, leading to more effective aerial object detection., Comment: ICIP 2024 Workshop accepted paper

A recent article in $\textit{Science}$ by Guess et al. estimated the effect of Facebook's news feed algorithm on exposure to misinformation and political information among Facebook users. However, its reporting and conclusions did not account for a series of temporary emergency changes to Facebook's news feed algorithm in the wake of the 2020 U.S. presidential election that were designed to diminish the spread of voter-fraud misinformation. Here, we demonstrate that these emergency measures systematically reduced the amount of misinformation in the control group of the study, which was using the news feed algorithm. This issue may have led readers to misinterpret the results of the study and to conclude that the Facebook news feed algorithm used outside of the study period mitigates political misinformation as compared to reverse chronological feed.

Training and prototype development in robot-assisted surgery requires appropriate and safe environments for the execution of surgical procedures. Current dry lab laparoscopy phantoms often lack the ability to mimic complex, interactive surgical tasks. This work presents an interactive surgical phantom for the cholecystectomy. The phantom enables the removal of the gallbladder during cholecystectomy by allowing manipulations and cutting interactions with the synthetic tissue. The force-displacement behavior of the gallbladder is modelled based on retraction demonstrations. The force model is compared to the force model of ex-vivo porcine gallbladders and evaluated on its ability to estimate retraction forces., Comment: As presented at CURAC 2023

Disordered iron germanium (FeGe) has recently garnered interest as a testbed for a variety of magnetic phenomena as well as for use in magnetic memory and logic applications. This is partially owing to its ability to host skyrmions and antiskyrmions -- nanoscale whirlpools of magnetic moments that could serve as information carriers in spintronic devices. In particular, a tunable skyrmion-antiskyrmion system may be created through precise control of the defect landscape in B20-phase FeGe, motivating developing methods to systematically tune disorder in this material and understand the ensuing structural properties. To this end, we investigate a route for modifying magnetic properties in FeGe. Specifically, we irradiate epitaxial B20-phase FeGe films with 2.8 MeV Au$^{4+}$ ions, which creates a dispersion of amorphized regions that may preferentially host antiskyrmions at densities controlled by the irradiation fluence. To further tune the disorder landscape, we conduct a systematic electron diffraction study with in-situ annealing, demonstrating the ability to recrystallize controllable fractions of the material at temperatures ranging from approximately 150$^{\circ}$ C to 250$^{\circ}$C. Finally, we describe the crystallization kinetics using the Johnson-Mehl-Avrami-Kolmogorov model, finding that the growth of crystalline grains is consistent with diffusion-controlled one-to-two dimensional growth with a decreasing nucleation rate.

This paper introduces StyleSpeech, a novel Text-to-Speech~(TTS) system that enhances the naturalness and accuracy of synthesized speech. Building upon existing TTS technologies, StyleSpeech incorporates a unique Style Decorator structure that enables deep learning models to simultaneously learn style and phoneme features, improving adaptability and efficiency through the principles of Lower Rank Adaptation~(LoRA). LoRA allows efficient adaptation of style features in pre-trained models. Additionally, we introduce a novel automatic evaluation metric, the LLM-Guided Mean Opinion Score (LLM-MOS), which employs large language models to offer an objective and robust protocol for automatically assessing TTS system performance. Extensive testing on benchmark datasets shows that our approach markedly outperforms existing state-of-the-art baseline methods in producing natural, accurate, and high-quality speech. These advancements not only pushes the boundaries of current TTS system capabilities, but also facilitate the application of TTS system in more dynamic and specialized, such as interactive virtual assistants, adaptive audiobooks, and customized voice for gaming. Speech samples can be found in https://style-speech.vercel.app

Polar nematic liquid crystals have two special features, compared with conventional nematic liquid crystals. First, because of flexoelectricity, the combination of polar order and splay reduces the free energy. Second, because of electrostatics, any splay generates a bound charge density, which increases the free energy. To assess the competition between these two effects, we develop a theory that combines flexoelectricity and electrostatics. The theory predicts a phase diagram that includes ferroelectric, antiferroelectric, and conventional nematic phases.

Predicting emotions elicited by news headlines can be challenging as the task is largely influenced by the varying nature of people's interpretations and backgrounds. Previous works have explored classifying discrete emotions directly from news headlines. We provide a different approach to tackling this problem by utilizing people's explanations of their emotion, written in free-text, on how they feel after reading a news headline. Using the dataset BU-NEmo+ (Gao et al., 2022), we found that for emotion classification, the free-text explanations have a strong correlation with the dominant emotion elicited by the headlines. The free-text explanations also contain more sentimental context than the news headlines alone and can serve as a better input to emotion classification models. Therefore, in this work we explored generating emotion explanations from headlines by training a sequence-to-sequence transformer model and by using pretrained large language model, ChatGPT (GPT-4). We then used the generated emotion explanations for emotion classification. In addition, we also experimented with training the pretrained T5 model for the intermediate task of explanation generation before fine-tuning it for emotion classification. Using McNemar's significance test, methods that incorporate GPT-generated free-text emotion explanations demonstrated significant improvement (P-value < 0.05) in emotion classification from headlines, compared to methods that only use headlines. This underscores the value of using intermediate free-text explanations for emotion prediction tasks with headlines., Comment: published at LREC-COLING 2024

Conventional intelligent systems based on deep neural network (DNN) models encounter challenges in achieving human-like continual learning due to catastrophic forgetting. Here, we propose a metaplasticity model inspired by human working memory, enabling DNNs to perform catastrophic forgetting-free continual learning without any pre- or post-processing. A key aspect of our approach involves implementing distinct types of synapses from stable to flexible, and randomly intermixing them to train synaptic connections with different degrees of flexibility. This strategy allowed the network to successfully learn a continuous stream of information, even under unexpected changes in input length. The model achieved a balanced tradeoff between memory capacity and performance without requiring additional training or structural modifications, dynamically allocating memory resources to retain both old and new information. Furthermore, the model demonstrated robustness against data poisoning attacks by selectively filtering out erroneous memories, leveraging the Hebb repetition effect to reinforce the retention of significant data., Comment: 25 pages, 5 figures, 1 table, 4 supplementary figures

Harmful suicide content on the Internet is a significant risk factor inducing suicidal thoughts and behaviors among vulnerable populations. Despite global efforts, existing resources are insufficient, specifically in high-risk regions like the Republic of Korea. Current research mainly focuses on understanding negative effects of such content or suicide risk in individuals, rather than on automatically detecting the harmfulness of content. To fill this gap, we introduce a harmful suicide content detection task for classifying online suicide content into five harmfulness levels. We develop a multi-modal benchmark and a task description document in collaboration with medical professionals, and leverage large language models (LLMs) to explore efficient methods for moderating such content. Our contributions include proposing a novel detection task, a multi-modal Korean benchmark with expert annotations, and suggesting strategies using LLMs to detect illegal and harmful content. Owing to the potential harm involved, we publicize our implementations and benchmark, incorporating an ethical verification process., Comment: 30 pages, 7 figures

Workloads in data processing clusters are often represented in the form of DAG (Directed Acyclic Graph) jobs. Scheduling DAG jobs is challenging. Simple heuristic scheduling algorithms are often adopted in practice in production data centres. There is much room for scheduling performance optimisation for cost saving. Recently, reinforcement learning approaches (like decima) have been attempted to optimise DAG job scheduling and demonstrate clear performance gain in comparison to traditional algorithms. However, reinforcement learning (RL) approaches face their own problems in real-world deployment. In particular, their black-box decision making processes and generalizability in unseen workloads may add a non-trivial burden to the cluster administrators. Moreover, adapting RL models on unseen workloads often requires significant amount of training data, which leaves edge cases run in a sub-optimal mode. To fill the gap, we propose a new method to distill a simple scheduling policy based on observations of the behaviours of a complex deep learning model. The simple model not only provides interpretability of scheduling decisions, but also adaptive to edge cases easily through tuning. We show that our method achieves high fidelity to the decisions made by deep learning models and outperforms these models when additional heuristics are taken into account., Comment: 20 pages, 18 figures

The brain prepares for learning even before interacting with the environment, by refining and optimizing its structures through spontaneous neural activity that resembles random noise. However, the mechanism of such a process has yet to be thoroughly understood, and it is unclear whether this process can benefit the algorithm of machine learning. Here, we study this issue using a neural network with a feedback alignment algorithm, demonstrating that pretraining neural networks with random noise increases the learning efficiency as well as generalization abilities without weight transport. First, we found that random noise training modifies forward weights to match backward synaptic feedback, which is necessary for teaching errors by feedback alignment. As a result, a network with pre-aligned weights learns notably faster than a network without random noise training, even reaching a convergence speed comparable to that of a backpropagation algorithm. Sequential training with both random noise and data brings weights closer to synaptic feedback than training solely with data, enabling more precise credit assignment and faster learning. We also found that each readout probability approaches the chance level and that the effective dimensionality of weights decreases in a network pretrained with random noise. This pre-regularization allows the network to learn simple solutions of a low rank, reducing the generalization loss during subsequent training. This also enables the network robustly to generalize a novel, out-of-distribution dataset. Lastly, we confirmed that random noise pretraining reduces the amount of meta-loss, enhancing the network ability to adapt to various tasks. Overall, our results suggest that random noise training with feedback alignment offers a straightforward yet effective method of pretraining that facilitates quick and reliable learning without weight transport.

This paper demonstrates the integration of Reinforcement Learning (RL) into quantum transpiling workflows, significantly enhancing the synthesis and routing of quantum circuits. By employing RL, we achieve near-optimal synthesis of Linear Function, Clifford, and Permutation circuits, up to 9, 11 and 65 qubits respectively, while being compatible with native device instruction sets and connectivity constraints, and orders of magnitude faster than optimization methods such as SAT solvers. We also achieve significant reductions in two-qubit gate depth and count for circuit routing up to 133 qubits with respect to other routing heuristics such as SABRE. We find the method to be efficient enough to be useful in practice in typical quantum transpiling pipelines. Our results set the stage for further AI-powered enhancements of quantum computing workflows.

Digital identity is evolving from centralized systems to a decentralized approach known as Self-Sovereign Identity (SSI). SSI empowers individuals to control their digital identities, eliminating reliance on third-party data custodians and reducing the risk of data breaches. However, the concept of trust in SSI remains complex and fragmented. This paper systematically analyzes trust in SSI in light of its components and threats posed by various actors in the system. As a result, we derive three distinct trust models that capture the threats and mitigations identified across SSI literature and implementations. Our work provides a foundational framework for future SSI research and development, including a comprehensive catalogue of SSI components and design requirements for trust, shortcomings in existing SSI systems and areas for further exploration., Comment: Accepted at PETS'24 Issue 3

We introduce a novel doubly-robust (DR) off-policy evaluation (OPE) estimator for Markov decision processes, DRUnknown, designed for situations where both the logging policy and the value function are unknown. The proposed estimator initially estimates the logging policy and then estimates the value function model by minimizing the asymptotic variance of the estimator while considering the estimating effect of the logging policy. When the logging policy model is correctly specified, DRUnknown achieves the smallest asymptotic variance within the class containing existing OPE estimators. When the value function model is also correctly specified, DRUnknown is optimal as its asymptotic variance reaches the semiparametric lower bound. We present experimental results conducted in contextual bandits and reinforcement learning to compare the performance of DRUnknown with that of existing methods.

The robustness of large language models (LLMs) becomes increasingly important as their use rapidly grows in a wide range of domains. Retrieval-Augmented Generation (RAG) is considered as a means to improve the trustworthiness of text generation from LLMs. However, how the outputs from RAG-based LLMs are affected by slightly different inputs is not well studied. In this work, we find that the insertion of even a short prefix to the prompt leads to the generation of outputs far away from factually correct answers. We systematically evaluate the effect of such prefixes on RAG by introducing a novel optimization technique called Gradient Guided Prompt Perturbation (GGPP). GGPP achieves a high success rate in steering outputs of RAG-based LLMs to targeted wrong answers. It can also cope with instructions in the prompts requesting to ignore irrelevant context. We also exploit LLMs' neuron activation difference between prompts with and without GGPP perturbations to give a method that improves the robustness of RAG-based LLMs through a highly effective detector trained on neuron activation triggered by GGPP generated prompts. Our evaluation on open-sourced LLMs demonstrates the effectiveness of our methods., Comment: 12 pages, 9 figures

The Incremental Potential Contact (IPC) method enables robust complex simulations of deformable objects with contact and friction. The key to IPC's robustness is its strict adherence to geometric constraints, avoiding intersections, which are a common cause of robustness issues in contact mechanics. A key element of the IPC approach to contact is a geometric barrier function, which is defined directly in the discrete setting. While IPC achieves its main goal of providing guarantees for contact constraints, its parameters need to be chosen carefully to avoid significant simulation artifacts and inaccuracies. We present a systematic derivation of an IPC-like continuum potential defined for smooth and piecewise smooth surfaces, starting from identifying a set of natural requirements for contact potentials, including the barrier property, locality, differentiable dependence of shape, and absence of forces in rest configurations, based on the idea of candidate sets. Our potential is formulated in a way independent of surface discretization. This new potential is suitable for piecewise-linear surfaces and its efficiency is similar to standard IPC. We demonstrate its behavior and compare it to IPC on a range of challenging contact examples.