Remove that Square Root: A New Efficient Scale-Invariant Version of AdaGrad

Adaptive methods are extremely popular in machine learning as they make learning rate tuning less expensive. This paper introduces a novel optimization algorithm named KATE, which presents a scale-invariant adaptation of the well-known AdaGrad algorithm. We prove the scale-invariance of KATE for the case of Generalized Linear Models. Moreover, for general smooth non-convex problems, we establish a convergence rate of $O \left(\frac{\log T}{\sqrt{T}} \right)$ for KATE, matching the best-known ones for AdaGrad and Adam. We also compare KATE to other state-of-the-art adaptive algorithms Adam and AdaGrad in numerical experiments with different problems, including complex machine learning tasks like image classification and text classification on real data. The results indicate that KATE consistently outperforms AdaGrad and matches/surpasses the performance of Adam in all considered scenarios.

Privacy-preserving Fine-tuning of Large Language Models through Flatness

The privacy concerns associated with the use of Large Language Models (LLMs) have grown recently with the development of LLMs such as ChatGPT. Differential Privacy (DP) techniques are explored in existing work to mitigate their privacy risks at the cost of generalization degradation. Our paper reveals that the flatness of DP-trained models' loss landscape plays an essential role in the trade-off between their privacy and generalization. We further propose a holistic framework to enforce appropriate weight flatness, which substantially improves model generalization with competitive privacy preservation. It innovates from three coarse-to-grained levels, including perturbation-aware min-max optimization on model weights within a layer, flatness-guided sparse prefix-tuning on weights across layers, and weight knowledge distillation between DP \& non-DP weights copies. Comprehensive experiments of both black-box and white-box scenarios are conducted to demonstrate the effectiveness of our proposal in enhancing generalization and maintaining DP characteristics. For instance, on text classification dataset QNLI, DP-Flat achieves similar performance with non-private full fine-tuning but with DP guarantee under privacy budget $\epsilon=3$, and even better performance given higher privacy budgets. Codes are provided in the supplement.

RADIA -- Radio Advertisement Detection with Intelligent Analytics

Radio advertising remains an integral part of modern marketing strategies, with its appeal and potential for targeted reach undeniably effective. However, the dynamic nature of radio airtime and the rising trend of multiple radio spots necessitates an efficient system for monitoring advertisement broadcasts. This study investigates a novel automated radio advertisement detection technique incorporating advanced speech recognition and text classification algorithms. RadIA's approach surpasses traditional methods by eliminating the need for prior knowledge of the broadcast content. This contribution allows for detecting impromptu and newly introduced advertisements, providing a comprehensive solution for advertisement detection in radio broadcasting. Experimental results show that the resulting model, trained on carefully segmented and tagged text data, achieves an F1-macro score of 87.76 against a theoretical maximum of 89.33. This paper provides insights into the choice of hyperparameters and their impact on the model's performance. This study demonstrates its potential to ensure compliance with advertising broadcast contracts and offer competitive surveillance. This groundbreaking research could fundamentally change how radio advertising is monitored and open new doors for marketing optimization.

SensoryT5: Infusing Sensorimotor Norms into T5 for Enhanced Fine-grained Emotion Classification

In traditional research approaches, sensory perception and emotion classification have traditionally been considered separate domains. Yet, the significant influence of sensory experiences on emotional responses is undeniable. The natural language processing (NLP) community has often missed the opportunity to merge sensory knowledge with emotion classification. To address this gap, we propose SensoryT5, a neuro-cognitive approach that integrates sensory information into the T5 (Text-to-Text Transfer Transformer) model, designed specifically for fine-grained emotion classification. This methodology incorporates sensory cues into the T5's attention mechanism, enabling a harmonious balance between contextual understanding and sensory awareness. The resulting model amplifies the richness of emotional representations. In rigorous tests across various detailed emotion classification datasets, SensoryT5 showcases improved performance, surpassing both the foundational T5 model and current state-of-the-art works. Notably, SensoryT5's success signifies a pivotal change in the NLP domain, highlighting the potential influence of neuro-cognitive data in refining machine learning models' emotional sensitivity.

Better (pseudo-)labels for semi-supervised instance segmentation

Despite the availability of large datasets for tasks like image classification and image-text alignment, labeled data for more complex recognition tasks, such as detection and segmentation, is less abundant. In particular, for instance segmentation annotations are time-consuming to produce, and the distribution of instances is often highly skewed across classes. While semi-supervised teacher-student distillation methods show promise in leveraging vast amounts of unlabeled data, they suffer from miscalibration, resulting in overconfidence in frequently represented classes and underconfidence in rarer ones. Additionally, these methods encounter difficulties in efficiently learning from a limited set of examples. We introduce a dual-strategy to enhance the teacher model's training process, substantially improving the performance on few-shot learning. Secondly, we propose a calibration correction mechanism that that enables the student model to correct the teacher's calibration errors. Using our approach, we observed marked improvements over a state-of-the-art supervised baseline performance on the LVIS dataset, with an increase of 2.8% in average precision (AP) and 10.3% gain in AP for rare classes.

Towards Efficient Active Learning in NLP via Pretrained Representations

Fine-tuning Large Language Models (LLMs) is now a common approach for text classification in a wide range of applications. When labeled documents are scarce, active learning helps save annotation efforts but requires retraining of massive models on each acquisition iteration. We drastically expedite this process by using pretrained representations of LLMs within the active learning loop and, once the desired amount of labeled data is acquired, fine-tuning that or even a different pretrained LLM on this labeled data to achieve the best performance. As verified on common text classification benchmarks with pretrained BERT and RoBERTa as the backbone, our strategy yields similar performance to fine-tuning all the way through the active learning loop but is orders of magnitude less computationally expensive. The data acquired with our procedure generalizes across pretrained networks, allowing flexibility in choosing the final model or updating it as newer versions get released.

Text Role Classification in Scientific Charts Using Multimodal Transformers

Text role classification involves classifying the semantic role of textual elements within scientific charts. For this task, we propose to finetune two pretrained multimodal document layout analysis models, LayoutLMv3 and UDOP, on chart datasets. The transformers utilize the three modalities of text, image, and layout as input. We further investigate whether data augmentation and balancing methods help the performance of the models. The models are evaluated on various chart datasets, and results show that LayoutLMv3 outperforms UDOP in all experiments. LayoutLMv3 achieves the highest F1-macro score of 82.87 on the ICPR22 test dataset, beating the best-performing model from the ICPR22 CHART-Infographics challenge. Moreover, the robustness of the models is tested on a synthetic noisy dataset ICPR22-N. Finally, the generalizability of the models is evaluated on three chart datasets, CHIME-R, DeGruyter, and EconBiz, for which we added labels for the text roles. Findings indicate that even in cases where there is limited training data, transformers can be used with the help of data augmentation and balancing methods. The source code and datasets are available on GitHub under https://github.com/hjkimk/text-role-classification

Quantum Mixed-State Self-Attention Network

The rapid advancement of quantum computing has increasingly highlighted its potential in the realm of machine learning, particularly in the context of natural language processing (NLP) tasks. Quantum machine learning (QML) leverages the unique capabilities of quantum computing to offer novel perspectives and methodologies for complex data processing and pattern recognition challenges. This paper introduces a novel Quantum Mixed-State Attention Network (QMSAN), which integrates the principles of quantum computing with classical machine learning algorithms, especially self-attention networks, to enhance the efficiency and effectiveness in handling NLP tasks. QMSAN model employs a quantum attention mechanism based on mixed states, enabling efficient direct estimation of similarity between queries and keys within the quantum domain, leading to more effective attention weight acquisition. Additionally, we propose an innovative quantum positional encoding scheme, implemented through fixed quantum gates within the quantum circuit, to enhance the model's accuracy. Experimental validation on various datasets demonstrates that QMSAN model outperforms existing quantum and classical models in text classification, achieving significant performance improvements. QMSAN model not only significantly reduces the number of parameters but also exceeds classical self-attention networks in performance, showcasing its strong capability in data representation and information extraction. Furthermore, our study investigates the model's robustness in different quantum noise environments, showing that QMSAN possesses commendable robustness to low noise.

Evaluating Named Entity Recognition: Comparative Analysis of Mono- and Multilingual Transformer Models on Brazilian Corporate Earnings Call Transcriptions

Named Entity Recognition (NER) is a Natural Language Processing technique for extracting information from textual documents. However, much of the existing research on NER has been centered around English-language documents, leaving a gap in the availability of datasets tailored to the financial domain in Portuguese. This study addresses the need for NER within the financial domain, focusing on Portuguese-language texts extracted from earnings call transcriptions of Brazilian banks. By curating a comprehensive dataset comprising 384 transcriptions and leveraging weak supervision techniques for annotation, we evaluate the performance of monolingual models trained on Portuguese (BERTimbau and PTT5) and multilingual models (mBERT and mT5). Notably, we introduce a novel approach that reframes the token classification task as a text generation problem, enabling fine-tuning and evaluation of T5 models. Following the fine-tuning of the models, we conduct an evaluation on the test dataset, employing performance and error metrics. Our findings reveal that BERT-based models consistently outperform T5-based models. Furthermore, while the multilingual models exhibit comparable macro F1-scores, BERTimbau demonstrates superior performance over PTT5. A manual analysis of sentences generated by PTT5 and mT5 unveils a degree of similarity ranging from 0.89 to 1.0, between the original and generated sentences. However, critical errors emerge as both models exhibit discrepancies, such as alterations to monetary and percentage values, underscoring the importance of accuracy and consistency in the financial domain. Despite these challenges, PTT5 and mT5 achieve impressive macro F1-scores of 98.52% and 98.85%, respectively, with our proposed approach. Furthermore, our study sheds light on notable disparities in memory and time consumption for inference across the models.

LLMLingua-2: Data Distillation for Efficient and Faithful Task-Agnostic Prompt Compression

This paper focuses on task-agnostic prompt compression for better generalizability and efficiency. Considering the redundancy in natural language, existing approaches compress prompts by removing tokens or lexical units according to their information entropy obtained from a causal language model such as LLaMa-7B. The challenge is that information entropy may be a suboptimal compression metric: (i) it only leverages unidirectional context and may fail to capture all essential information needed for prompt compression; (ii) it is not aligned with the prompt compression objective. To address these issues, we propose a data distillation procedure to derive knowledge from an LLM to compress prompts without losing crucial information, and meantime, introduce an extractive text compression dataset. We formulate prompt compression as a token classification problem to guarantee the faithfulness of the compressed prompt to the original one, and use a Transformer encoder as the base architecture to capture all essential information for prompt compression from the full bidirectional context. Our approach leads to lower latency by explicitly learning the compression objective with smaller models such as XLM-RoBERTa-large and mBERT. We evaluate our method on both in-domain and out-of-domain datasets, including MeetingBank, LongBench, ZeroScrolls, GSM8K, and BBH. Despite its small size, our model shows significant performance gains over strong baselines and demonstrates robust generalization ability across different LLMs. Additionally, our model is 3x-6x faster than existing prompt compression methods, while accelerating the end-to-end latency by 1.6x-2.9x with compression ratios of 2x-5x.