We introduce the LiLiuM series of large language models (LLMs): 1B, 7B, and 13B parameter models developed 100% in-house to fit eBay's specific needs in the e-commerce domain. This gives eBay full control over all aspects of the models including license, data, vocabulary, and architecture. We expect these models to be used as a foundation for fine-tuning and instruction-tuning, eliminating dependencies to external models. The LiLiuM LLMs have been trained on 3 trillion tokens of multilingual text from general and e-commerce domain. They perform similar to the popular LLaMA-2 models on English natural language understanding (NLU) benchmarks. At the same time, we outperform LLaMA-2 on non-English NLU tasks, machine translation and on e-commerce specific downstream tasks. As part of our data mixture, we utilize the newly released RedPajama-V2 dataset for training and share our insights regarding data filtering and deduplication. We also discuss in detail how to serialize structured data for use in autoregressive language modeling. We provide insights on the effects of including code and parallel machine translation data in pre-training. Furthermore, we develop our own tokenizer and model vocabulary, customized towards e-commerce. This way, we can achieve up to 34% speed-up in text generation on eBay-specific downstream tasks compared to LLaMA-2. Finally, in relation to LLM pretraining, we show that checkpoint averaging can further improve over the best individual model checkpoint.

Deep learning (DL) models have emerged as a powerful tool in avian bioacoustics to assess environmental health. To maximize the potential of cost-effective and minimal-invasive passive acoustic monitoring (PAM), DL models must analyze bird vocalizations across a wide range of species and environmental conditions. However, data fragmentation challenges a comprehensive evaluation of generalization performance. Therefore, we introduce the BirdSet dataset, comprising approximately 520,000 global bird recordings for training and over 400 hours of PAM recordings for testing. Our benchmark offers baselines for several DL models to enhance comparability and consolidate research across studies, along with code implementations that include comprehensive training and evaluation protocols.

Code-switching is a widely prevalent linguistic phenomenon in multilingual societies like India. Building speech-to-text models for code-switched speech is challenging due to limited availability of datasets. In this work, we focus on the problem of spoken translation (ST) of code-switched speech in Indian languages to English text. We present a new end-to-end model architecture COSTA that scaffolds on pretrained automatic speech recognition (ASR) and machine translation (MT) modules (that are more widely available for many languages). Speech and ASR text representations are fused using an aligned interleaving scheme and are fed further as input to a pretrained MT module; the whole pipeline is then trained end-to-end for spoken translation using synthetically created ST data. We also release a new evaluation benchmark for code-switched Bengali-English, Hindi-English, Marathi-English and Telugu- English speech to English text. COSTA significantly outperforms many competitive cascaded and end-to-end multimodal baselines by up to 3.5 BLEU points.

Continual Test-Time Adaptation (TTA) seeks to adapt a source pre-trained model to continually changing, unlabeled target domains. Existing TTA methods are typically designed for environments where domain changes occur gradually and can struggle in more dynamic scenarios. Inspired by the principles of online K-Means, this paper introduces a novel approach to continual TTA through visual prompting. We propose a Dynamic Prompt Coreset that not only preserves knowledge from previously visited domains but also accommodates learning from new potential domains. This is complemented by a distance-based weight updating mechanism that ensures the coreset remains current and relevant. Our approach employs a fixed model architecture alongside the coreset and an innovative updating system to effectively mitigate challenges such as catastrophic forgetting and error accumulation. Extensive testing across various benchmarks-including ImageNet-C, CIFAR100-C, and CIFAR10-C-demonstrates that our method consistently outperforms state-of-the-art (SOTA) alternatives, particularly excelling in dynamically changing environments.

The intersection between poetry and music provides an interesting case for computational creativity, yet remains relatively unexplored. This paper explores the integration of poetry and music through the lens of beat patterns, investigating whether a byte-based language model can generate words that fit specific beat patterns within the context of poetry. Drawing on earlier studies, we developed a method to train a byte-based transformer model, ByT5, to align poems with beat patterns. The results demonstrate a high level of beat alignment while maintaining semantic coherence. Future work will aim to improve the model's ability to create complete beat-aligned poems.

We revisit the INTERSPEECH 2009 Emotion Challenge -- the first ever speech emotion recognition (SER) challenge -- and evaluate a series of deep learning models that are representative of the major advances in SER research in the time since then. We start by training each model using a fixed set of hyperparameters, and further fine-tune the best-performing models of that initial setup with a grid search. Results are always reported on the official test set with a separate validation set only used for early stopping. Most models score below or close to the official baseline, while they marginally outperform the original challenge winners after hyperparameter tuning. Our work illustrates that, despite recent progress, FAU-AIBO remains a very challenging benchmark. An interesting corollary is that newer methods do not consistently outperform older ones, showing that progress towards `solving' SER is not necessarily monotonic.

This study investigates the capabilities of Large Language Models (LLMs), specifically GPT-4, in the context of Binary Reverse Engineering (RE). Employing a structured experimental approach, we analyzed the LLM's performance in interpreting and explaining human-written and decompiled codes. The research encompassed two phases: the first on basic code interpretation and the second on more complex malware analysis. Key findings indicate LLMs' proficiency in general code understanding, with varying effectiveness in detailed technical and security analyses. The study underscores the potential and current limitations of LLMs in reverse engineering, revealing crucial insights for future applications and improvements. Also, we examined our experimental methodologies, such as methods of evaluation and data constraints, which provided us with a technical vision for any future research activity in this field.

When new languages need to be integrated into a multilingual ASR system, a naive Recent years have witnessed significant progress in multilingual approach is to fine-tune the ASR model using data from these automatic speech recognition (ASR), driven by the emergence new languages. Unfortunately, this often results in catastrophic of end-to-end (E2E) models and the scaling of multilingual forgetting, referring to the phenomenon that the recognition performance datasets. Despite that, two main challenges persist in multilingual of base languages tends to decline. To solve the above ASR: language interference and the incorporation of problem, Li et al. [26] proposes lifelong learning [27] solution new languages without degrading the performance of the existing which remedies the language interference problem by mixing ones. This paper proposes LoRA-Whisper, which incorporates base language data and new language data. However, this approach LoRA matrix into Whisper for multilingual ASR, is inefficient and time-consuming. Libera et al. [28] explores effectively mitigating language interference. Furthermore, by various continual learning methods [29-34] to address leveraging LoRA and the similarities between languages, we the issue of catastrophic forgetting. While these approaches can achieve better performance on new languages while upholding have helped alleviate the problem, it still persists.

Randomized Smoothing (RS) is currently a scalable certified defense method providing robustness certification against adversarial examples. Although significant progress has been achieved in providing defenses against $\ell_p$ adversaries, the interaction between the smoothing distribution and the robustness certification still remains vague. In this work, we comprehensively study the effect of two families of distributions, named Exponential Standard Gaussian (ESG) and Exponential General Gaussian (EGG) distributions, on Randomized Smoothing and Double Sampling Randomized Smoothing (DSRS). We derive an analytic formula for ESG's certified radius, which converges to the origin formula of RS as the dimension $d$ increases. Additionally, we prove that EGG can provide tighter constant factors than DSRS in providing $\Omega(\sqrt{d})$ lower bounds of $\ell_2$ certified radius, and thus further addresses the curse of dimensionality in RS. Our experiments on real-world datasets confirm our theoretical analysis of the ESG distributions, that they provide almost the same certification under different exponents $\eta$ for both RS and DSRS. In addition, EGG brings a significant improvement to the DSRS certification, but the mechanism can be different when the classifier properties are different. Compared to the primitive DSRS, the increase in certified accuracy provided by EGG is prominent, up to 6.4% on ImageNet.

Deriving meaningful representations from complex, high-dimensional data in unsupervised settings is crucial across diverse machine learning applications. This paper introduces a framework for multi-scale graph network embedding based on spectral graph wavelets that employs a contrastive learning approach. A significant feature of the proposed embedding is its capacity to establish a correspondence between the embedding space and the input feature space which aids in deriving feature importance of the original features. We theoretically justify our approach and demonstrate that, in Paley-Wiener spaces on combinatorial graphs, the spectral graph wavelets operator offers greater flexibility and better control over smoothness properties compared to the Laplacian operator. We validate the effectiveness of our proposed graph embedding on a variety of public datasets through a range of downstream tasks, including clustering and unsupervised feature importance.