Ulf Krumnack

Mohamad Ballout, Ulf Krumnack, Gunther Heidemann et al.

Pre-trained language models have recently emerged as a powerful tool for fine-tuning a variety of language tasks. Ideally, when models are pre-trained on large amount of data, they are expected to gain implicit knowledge. In this paper, we investigate the ability of pre-trained language models to generalize to different non-language tasks. In particular, we test them on tasks from different domains such as computer vision, reasoning on hierarchical data, and protein fold prediction. The four pre-trained models that we used, T5, BART, BERT, and GPT-2 achieve outstanding results. They all have similar performance and they outperform transformers that are trained from scratch by a large margin. For instance, pre-trained language models perform better on the Listops dataset, with an average accuracy of 58.7\%, compared to transformers trained from scratch, which have an average accuracy of 29.0\%. The significant improvement demonstrated across three types of datasets suggests that pre-training on language helps the models to acquire general knowledge, bringing us a step closer to general AI. We also showed that reducing the number of parameters in pre-trained language models does not have a great impact as the performance drops slightly when using T5-Small instead of T5-Base. In fact, when using only 2\% of the parameters, we achieved a great improvement compared to training from scratch. Finally, in contrast to prior work, we find out that using pre-trained embeddings for the input layer is necessary to achieve the desired results.

Mohamad Ballout, Ulf Krumnack, Gunther Heidemann et al.

Enhancing small language models for real-life application deployment is a significant challenge facing the research community. Due to the difficulties and costs of using large language models, researchers are seeking ways to effectively deploy task-specific small models. In this work, we introduce a simple yet effective knowledge distillation method to improve the performance of small language models. Our approach utilizes a teacher model with approximately 3 billion parameters to identify the most influential tokens in its decision-making process. These tokens are extracted from the input based on their attribution scores relative to the output, using methods like saliency maps. These important tokens are then provided as rationales to a student model, aiming to distill the knowledge of the teacher model. This method has proven to be effective, as demonstrated by testing it on four diverse datasets, where it shows improvement over both standard fine-tuning methods and state-of-the-art knowledge distillation models. Furthermore, we explore explanations of the success of the model by analyzing the important tokens extracted from the teacher model. Our findings reveal that in 68\% of cases, specifically in datasets where labels are part of the answer, such as multiple-choice questions, the extracted tokens are part of the ground truth.

Mohamad Ballout, Ulf Krumnack, Gunther Heidemann et al.

Investigating deep learning language models has always been a significant research area due to the ``black box" nature of most advanced models. With the recent advancements in pre-trained language models based on transformers and their increasing integration into daily life, addressing this issue has become more pressing. In order to achieve an explainable AI model, it is essential to comprehend the procedural steps involved and compare them with human thought processes. Thus, in this paper, we use simple, well-understood non-language tasks to explore these models' inner workings. Specifically, we apply a pre-trained language model to constrained arithmetic problems with hierarchical structure, to analyze their attention weight scores and hidden states. The investigation reveals promising results, with the model addressing hierarchical problems in a moderately structured manner, similar to human problem-solving strategies. Additionally, by inspecting the attention weights layer by layer, we uncover an unconventional finding that layer 10, rather than the model's final layer, is the optimal layer to unfreeze for the least parameter-intensive approach to fine-tune the model. We support these findings with entropy analysis and token embeddings similarity analysis. The attention analysis allows us to hypothesize that the model can generalize to longer sequences in ListOps dataset, a conclusion later confirmed through testing on sequences longer than those in the training set. Lastly, by utilizing a straightforward task in which the model predicts the winner of a Tic Tac Toe game, we identify limitations in attention analysis, particularly its inability to capture 2D patterns.

Yusuf Brima, Ulf Krumnack, Simone Pika et al.

Disentangled representation learning in speech processing has lagged behind other domains, largely due to the lack of datasets with annotated generative factors for robust evaluation. To address this, we propose SynSpeech, a novel large-scale synthetic speech dataset specifically designed to enable research on disentangled speech representations. SynSpeech includes controlled variations in speaker identity, spoken text, and speaking style, with three dataset versions to support experimentation at different levels of complexity. In this study, we present a comprehensive framework to evaluate disentangled representation learning techniques, applying both linear probing and established supervised disentanglement metrics to assess the modularity, compactness, and informativeness of the representations learned by a state-of-the-art model. Using the RAVE model as a test case, we find that SynSpeech facilitates benchmarking across a range of factors, achieving promising disentanglement of simpler features like gender and speaking style, while highlighting challenges in isolating complex attributes like speaker identity. This benchmark dataset and evaluation framework fills a critical gap, supporting the development of more robust and interpretable speech representation learning methods.

Tom Pieper, Mohamad Ballout, Ulf Krumnack et al.

This paper explores the enhancement of small language models through strategic dataset augmentation via ChatGPT-3.5-Turbo, in the domain of Natural Language Inference (NLI). By employing knowledge distillation-based techniques and synthetic dataset augmentation, we aim to bridge the performance gap between large language models (LLMs) and small language models (SLMs) without the immense cost of human annotation. Our methods involve two forms of rationale generation--information extraction and informed reasoning--to enrich the ANLI dataset. We then fine-tune T5-Small on these augmented datasets, evaluating its performance against an established benchmark. Our findings reveal that the incorporation of synthetic rationales significantly improves the model's ability to comprehend natural language, leading to 1.3\% and 2.3\% higher classification accuracy, respectively, on the ANLI dataset, demonstrating the potential of leveraging LLMs for dataset augmentation. This approach not only enhances the performance of smaller models on complex tasks but also introduces a cost-effective method for fine-tuning smaller language models. By advancing our understanding of knowledge distillation and fine-tuning strategies, this work contributes to the ongoing effort to create more capable and efficient NLP systems.

Yusuf Brima, Ulf Krumnack, Simone Pika et al.

Self-supervised learning (SSL) has emerged as a promising paradigm for learning flexible speech representations from unlabeled data. By designing pretext tasks that exploit statistical regularities, SSL models can capture useful representations that are transferable to downstream tasks. This study provides an empirical analysis of Barlow Twins (BT), an SSL technique inspired by theories of redundancy reduction in human perception. On downstream tasks, BT representations accelerated learning and transferred across domains. However, limitations exist in disentangling key explanatory factors, with redundancy reduction and invariance alone insufficient for factorization of learned latents into modular, compact, and informative codes. Our ablations study isolated gains from invariance constraints, but the gains were context-dependent. Overall, this work substantiates the potential of Barlow Twins for sample-efficient speech encoding. However, challenges remain in achieving fully hierarchical representations. The analysis methodology and insights pave a path for extensions incorporating further inductive priors and perceptual principles to further enhance the BT self-supervision framework.

Mohamad Ballout, Ulf Krumnack, Gunther Heidemann et al.

Our research demonstrates the significant benefits of using fine-tuning with explanations to enhance the performance of language models. Unlike prompting, which maintains the model's parameters, fine-tuning allows the model to learn and update its parameters during a training phase. In this study, we applied fine-tuning to various sized language models using data that contained explanations of the output rather than merely presenting the answers. We found that even smaller language models with as few as 60 million parameters benefited substantially from this approach. Interestingly, our results indicated that the detailed explanations were more beneficial to smaller models than larger ones, with the latter gaining nearly the same advantage from any form of explanation, irrespective of its length. Additionally, we demonstrate that the inclusion of explanations enables the models to solve tasks that they were not able to solve without explanations. Lastly, we argue that despite the challenging nature of adding explanations, samples that contain explanations not only reduce the volume of data required for training but also promote a more effective generalization by the model. In essence, our findings suggest that fine-tuning with explanations significantly bolsters the performance of large language models.

Yusuf Brima, Ulf Krumnack, Simone Pika et al.

This paper tackles the scarcity of benchmarking data in disentangled auditory representation learning. We introduce SynTone, a synthetic dataset with explicit ground truth explanatory factors for evaluating disentanglement techniques. Benchmarking state-of-the-art methods on SynTone highlights its utility for method evaluation. Our results underscore strengths and limitations in audio disentanglement, motivating future research.

Mats L. Richter, Julius Schöning, Anna Wiedenroth et al.

When optimizing convolutional neural networks (CNN) for a specific image-based task, specialists commonly overshoot the number of convolutional layers in their designs. By implication, these CNNs are unnecessarily resource intensive to train and deploy, with diminishing beneficial effects on the predictive performance. The features a convolutional layer can process are strictly limited by its receptive field. By layer-wise analyzing the size of the receptive fields, we can reliably predict sequences of layers that will not contribute qualitatively to the test accuracy in the given CNN architecture. Based on this analysis, we propose design strategies based on a so-called border layer. This layer allows to identify unproductive convolutional layers and hence to resolve these inefficiencies, optimize the explainability and the computational performance of CNNs. Since neither the strategies nor the analysis requires training of the actual model, these insights allow for a very efficient design process of CNN architectures, which might be automated in the future.

Mats L. Richter, Leila Malihi, Anne-Kathrin Patricia Windler et al.

In this work we explore the information processing inside neural networks using logistic regression probes \cite{probes} and the saturation metric \cite{featurespace_saturation}. We show that problem difficulty and neural network capacity affect the predictive performance in an antagonistic manner, opening the possibility of detecting over- and under-parameterization of neural networks for a given task. We further show that the observed effects are independent from previously reported pathological patterns like the ``tail pattern'' described in \cite{featurespace_saturation}. Finally we are able to show that saturation patterns converge early during training, allowing for a quicker cycle time during analysis

Mats L. Richter, Wolf Byttner, Ulf Krumnack et al.

Fully convolutional neural networks can process input of arbitrary size by applying a combination of downsampling and pooling. However, we find that fully convolutional image classifiers are not agnostic to the input size but rather show significant differences in performance: presenting the same image at different scales can result in different outcomes. A closer look reveals that there is no simple relationship between input size and model performance (no `bigger is better'), but that each each network has a preferred input size, for which it shows best results. We investigate this phenomenon by applying different methods, including spectral analysis of layer activations and probe classifiers, showing that there are characteristic features depending on the network architecture. From this we find that the size of discriminatory features is critically influencing how the inference process is distributed among the layers.