Pretrained language models (PLMs) are key components in NLP, but they contain strong social biases. Quantifying these biases is challenging because current methods focusing on fill-the-mask objectives are sensitive to slight changes in input. To address this, we propose a bias probing technique called LABDet, for evaluating social bias in PLMs with a robust and language-agnostic method. For nationality as a case study, we show that LABDet `surfaces' nationality bias by training a classifier on top of a frozen PLM on non-nationality sentiment detection. We find consistent patterns of nationality bias across monolingual PLMs in six languages that align with historical and political context. We also show for English BERT that bias surfaced by LABDet correlates well with bias in the pretraining data; thus, our work is one of the few studies that directly links pretraining data to PLM behavior. Finally, we verify LABDet's reliability and applicability to different templates and languages through an extensive set of robustness checks. We publicly share our code and dataset in https://github.com/akoksal/LABDet.

With the bomb ignited by ChatGPT, Transformer-based Large Language Models (LLMs) have paved a revolutionary path toward Artificial General Intelligence (AGI) and have been applied in diverse areas as knowledge bases, human interfaces, and dynamic agents. However, a prevailing limitation exists: many current LLMs, constrained by resources, are primarily pre-trained on shorter texts, rendering them less effective for longer-context prompts, commonly encountered in real-world settings. In this paper, we present a comprehensive survey focusing on the advancement of model architecture in Transformer-based LLMs to optimize long-context capabilities across all stages from pre-training to inference. We firstly delineate and analyze the problems of handling long-context input and output with the current Transformer-based models. Then, we mainly offer a holistic taxonomy to navigate the landscape of Transformer upgrades on architecture to solve these problems. Afterward, we provide the investigation on wildly used evaluation necessities tailored for long-context LLMs, including datasets, metrics, and baseline models, as well as some amazing optimization toolkits like libraries, systems, and compilers to augment LLMs' efficiency and efficacy across different stages. Finally, we further discuss the predominant challenges and potential avenues for future research in this domain. Additionally, we have established a repository where we curate relevant literature with real-time updates at https://github.com/Strivin0311/long-llms-learning.

Italy is characterized by a one-of-a-kind linguistic diversity landscape in Europe, which implicitly encodes local knowledge, cultural traditions, artistic expressions and history of its speakers. However, most local languages and dialects in Italy are at risk of disappearing within few generations. The NLP community has recently begun to engage with endangered languages, including those of Italy. Yet, most efforts assume that these varieties are under-resourced language monoliths with an established written form and homogeneous functions and needs, and thus highly interchangeable with each other and with high-resource, standardized languages. In this paper, we introduce the linguistic context of Italy and challenge the default machine-centric assumptions of NLP for Italy's language varieties. We advocate for a shift in the paradigm from machine-centric to speaker-centric NLP, and provide recommendations and opportunities for work that prioritizes languages and their speakers over technological advances. To facilitate the process, we finally propose building a local community towards responsible, participatory efforts aimed at supporting vitality of languages and dialects of Italy.

In this report, we delve into two critical research inquiries. Firstly, we explore the extent to which Reinforcement Learning (RL) agents exhibit multimodal distributions in the context of stop-and-go traffic scenarios. Secondly, we investigate how RL-controlled Robot Vehicles (RVs) effectively navigate their direction and coordinate with other vehicles in complex traffic environments. Our analysis encompasses an examination of multimodality within queue length, outflow, and platoon size distributions for both Robot and Human-driven Vehicles (HVs). Additionally, we assess the Pearson coefficient correlation, shedding light on relationships between queue length and outflow, considering both identical and differing travel directions. Furthermore, we delve into causal inference models, shedding light on the factors influencing queue length across scenarios involving varying travel directions. Through these investigations, this report contributes valuable insights into the behaviors of mixed traffic (RVs and HVs) in traffic management and coordination.

Software bots have attracted increasing interest and popularity in both research and society. Their contributions span automation, digital twins, game characters with conscious-like behavior, and social media. However, there is still a lack of intelligent bots that can adapt to the variability and dynamic nature of digital web environments. Unlike human users, they have difficulty understanding and exploiting the affordances across multiple virtual environments. Despite the hype, bots with human user-like cognition do not currently exist. Chatbots, for instance, lack situational awareness on the digital platforms where they operate, preventing them from enacting meaningful and autonomous intelligent behavior similar to human users. In this survey, we aim to explore the role of cognitive architectures in supporting efforts towards engineering software bots with advanced general intelligence. We discuss how cognitive architectures can contribute to creating intelligent software bots. Furthermore, we highlight key architectural recommendations for the future development of autonomous, user-like cognitive bots.

Automated disease, weed and crop classification with computer vision will be invaluable in the future of agriculture. However, existing model architectures like ResNet, EfficientNet and ConvNeXt often underperform on smaller, specialised datasets typical of such projects. We address this gap with informed data collection and the development of a new CNN architecture, PhytNet. Utilising a novel dataset of infrared cocoa tree images, we demonstrate PhytNet's development and compare its performance with existing architectures. Data collection was informed by analysis of spectroscopy data, which provided useful insights into the spectral characteristics of cocoa trees. Such information could inform future data collection and model development. Cocoa was chosen as a focal species due to the diverse pathology of its diseases, which pose significant challenges for detection. ResNet18 showed some signs of overfitting, while EfficientNet variants showed distinct signs of overfitting. By contrast, PhytNet displayed excellent attention to relevant features, no overfitting, and an exceptionally low computation cost (1.19 GFLOPS). As such PhytNet is a promising candidate for rapid disease or plant classification, or precise localisation of disease symptoms for autonomous systems.

Lip segmentation is crucial in computer vision, especially for lip reading. Despite extensive face segmentation research, lip segmentation has received limited attention. The aim of this study is to compare state-of-the-art lip segmentation models using a standardized setting and a publicly available dataset. Five techniques, namely EHANet, Mask2Former, BiSeNet V2, PIDNet, and STDC1, are qualitatively selected based on their reported performance, inference time, code availability, recency, and popularity. The CelebAMask-HQ dataset, comprising manually annotated face images, is used to fairly assess the lip segmentation performance of the selected models. Inference experiments are conducted on a Raspberry Pi4 to emulate limited computational resources. The results show that Mask2Former and EHANet have the best performances in terms of mIoU score. BiSeNet V2 demonstrate competitive performance, while PIDNet excels in recall but has lower precision. Most models present inference time ranging from 1000 to around 3000 milliseconds on a Raspberry Pi4, with PIDNet having the lowest mean inference time. This study provides a comprehensive evaluation of lip segmentation models, highlighting their performance and inference times. The findings contribute to the development of lightweight techniques and establish benchmarks for future advances in lip segmentation, especially in IoT and edge computing scenarios.

People naturally understand emotions, thus permitting a machine to do the same could open new paths for human-computer interaction. Facial expressions can be very useful for emotion recognition techniques, as these are the biggest transmitters of non-verbal cues capable of being correlated with emotions. Several techniques are based on Convolutional Neural Networks (CNNs) to extract information in a machine learning process. However, simple CNNs are not always sufficient to locate points of interest on the face that can be correlated with emotions. In this work, we intend to expand the capacity of emotion recognition techniques by proposing the usage of Facial Action Units (AUs) recognition techniques to recognize emotions. This recognition will be based on the Facial Action Coding System (FACS) and computed by a machine learning system. In particular, our method expands over EmotiRAM, an approach for multi-cue emotion recognition, in which we improve over their facial encoding module.

Large neural networks pretrained on web-scale corpora are central to modern machine learning. In this paradigm, the distribution of the large, heterogeneous pretraining data rarely matches that of the application domain. This work considers modifying the pretraining distribution in the case where one has a small sample of data reflecting the targeted test conditions. We propose an algorithm motivated by a recent formulation of this setting as an online, bilevel optimization problem. With scalability in mind, our algorithm prioritizes computing gradients at training points which are likely to most improve the loss on the targeted distribution. Empirically, we show that in some cases this approach is beneficial over existing strategies from the domain adaptation literature but may not succeed in other cases. We propose a simple test to evaluate when our approach can be expected to work well and point towards further research to address current limitations.

The task of multispectral image segmentation (segmentation of images with numerous channels/bands, each capturing a specific range of wavelengths of electromagnetic radiation) has been previously explored in contexts with large amounts of labeled data. However, these models tend not to generalize well to datasets of smaller size. In this paper, we propose a novel approach for improving few-shot segmentation performance on multispectral images using reinforcement learning to generate representations. These representations are generated in the form of mathematical expressions between channels and are tailored to the specific class being segmented. Our methodology involves training an agent to identify the most informative expressions, updating the dataset using these expressions, and then using the updated dataset to perform segmentation. Due to the limited length of the expressions, the model receives useful representations without any added risk of overfitting. We evaluate the effectiveness of our approach on several multispectral datasets and demonstrate its effectiveness in boosting the performance of segmentation algorithms.