Can Large Language Models Learn Independent Causal Mechanisms?

Despite impressive performance on language modelling and complex reasoning tasks, Large Language Models (LLMs) fall short on the same tasks in uncommon settings or with distribution shifts, exhibiting some lack of generalisation ability. This issue has usually been alleviated by feeding more training data into the LLM. However, this method is brittle, as the scope of tasks may not be readily predictable or may evolve, and updating the model with new data generally requires extensive additional training. By contrast, systems, such as causal models, that learn abstract variables and causal relationships can demonstrate increased robustness against changes in the distribution. One reason for this success is the existence and use of Independent Causal Mechanisms (ICMs) representing high-level concepts that only sparsely interact. In this work, we apply two concepts from causality to learn ICMs within LLMs. We develop a new LLM architecture composed of multiple sparsely interacting language modelling modules. We introduce a routing scheme to induce specialisation of the network into domain-specific modules. We also present a Mutual Information minimisation objective that trains a separate module to learn abstraction and domain-invariant mechanisms. We show that such causal constraints can improve out-of-distribution performance on abstract and causal reasoning tasks.

Behaviour Modelling of Social Animals via Causal Structure Discovery and Graph Neural Networks

Better understanding the natural world is a crucial task with a wide range of applications. In environments with close proximity between humans and animals, such as zoos, it is essential to better understand the causes behind animal behaviour and what interventions are responsible for changes in their behaviours. This can help to predict unusual behaviours, mitigate detrimental effects and increase the well-being of animals. There has been work on modelling the dynamics behind swarms of birds and insects but the complex social behaviours of mammalian groups remain less explored. In this work, we propose a method to build behavioural models using causal structure discovery and graph neural networks for time series. We apply this method to a mob of meerkats in a zoo environment and study its ability to predict future actions and model the behaviour distribution at an individual-level and at a group level. We show that our method can match and outperform standard deep learning architectures and generate more realistic data, while using fewer parameters and providing increased interpretability.

Tackling Bias in Pre-trained Language Models: Current Trends and Under-represented Societies

The benefits and capabilities of pre-trained language models (LLMs) in current and future innovations are vital to any society. However, introducing and using LLMs comes with biases and discrimination, resulting in concerns about equality, diversity and fairness, and must be addressed. While understanding and acknowledging bias in LLMs and developing mitigation strategies are crucial, the generalised assumptions towards societal needs can result in disadvantages towards under-represented societies and indigenous populations. Furthermore, the ongoing changes to actual and proposed amendments to regulations and laws worldwide also impact research capabilities in tackling the bias problem. This research presents a comprehensive survey synthesising the current trends and limitations in techniques used for identifying and mitigating bias in LLMs, where the overview of methods for tackling bias are grouped into metrics, benchmark datasets, and mitigation strategies. The importance and novelty of this survey are that it explores the perspective of under-represented societies. We argue that current practices tackling the bias problem cannot simply be 'plugged in' to address the needs of under-represented societies. We use examples from New Zealand to present requirements for adopting existing techniques to under-represented societies.

Poison is Not Traceless: Fully-Agnostic Detection of Poisoning Attacks

The performance of machine learning models depends on the quality of the underlying data. Malicious actors can attack the model by poisoning the training data. Current detectors are tied to either specific data types, models, or attacks, and therefore have limited applicability in real-world scenarios. This paper presents a novel fully-agnostic framework, DIVA (Detecting InVisible Attacks), that detects attacks solely relying on analyzing the potentially poisoned data set. DIVA is based on the idea that poisoning attacks can be detected by comparing the classifier's accuracy on poisoned and clean data and pre-trains a meta-learner using Complexity Measures to estimate the otherwise unknown accuracy on a hypothetical clean dataset. The framework applies to generic poisoning attacks. For evaluation purposes, in this paper, we test DIVA on label-flipping attacks.

Fast Adversarial Label-Flipping Attack on Tabular Data

Machine learning models are increasingly used in fields that require high reliability such as cybersecurity. However, these models remain vulnerable to various attacks, among which the adversarial label-flipping attack poses significant threats. In label-flipping attacks, the adversary maliciously flips a portion of training labels to compromise the machine learning model. This paper raises significant concerns as these attacks can camouflage a highly skewed dataset as an easily solvable classification problem, often misleading machine learning practitioners into lower defenses and miscalculations of potential risks. This concern amplifies in tabular data settings, where identifying true labels requires expertise, allowing malicious label-flipping attacks to easily slip under the radar. To demonstrate this risk is inherited in the adversary's objective, we propose FALFA (Fast Adversarial Label-Flipping Attack), a novel efficient attack for crafting adversarial labels. FALFA is based on transforming the adversary's objective and employs linear programming to reduce computational complexity. Using ten real-world tabular datasets, we demonstrate FALFA's superior attack potential, highlighting the need for robust defenses against such threats.

Challenges in Annotating Datasets to Quantify Bias in Under-represented Society

Recent advances in artificial intelligence, including the development of highly sophisticated large language models (LLM), have proven beneficial in many real-world applications. However, evidence of inherent bias encoded in these LLMs has raised concerns about equity. In response, there has been an increase in research dealing with bias, including studies focusing on quantifying bias and developing debiasing techniques. Benchmark bias datasets have also been developed for binary gender classification and ethical/racial considerations, focusing predominantly on American demographics. However, there is minimal research in understanding and quantifying bias related to under-represented societies. Motivated by the lack of annotated datasets for quantifying bias in under-represented societies, we endeavoured to create benchmark datasets for the New Zealand (NZ) population. We faced many challenges in this process, despite the availability of three annotators. This research outlines the manual annotation process, provides an overview of the challenges we encountered and lessons learnt, and presents recommendations for future research.

Large Language Models Are Not Abstract Reasoners

Large Language Models have shown tremendous performance on a large variety of natural language processing tasks, ranging from text comprehension to common sense reasoning. However, the mechanisms responsible for this success remain unknown, and it is unclear whether LLMs can achieve human-like cognitive capabilities or whether these models are still fundamentally limited. Abstract reasoning is a fundamental task for cognition, consisting of finding and applying a general pattern from few data. Evaluating deep neural architectures on this task could give insight into their potential limitations regarding reasoning and their broad generalisation abilities, yet this is currently an under-explored area. In this paper, we perform extensive evaluations of state-of-the-art LLMs on abstract reasoning tasks, showing that they achieve very limited performance in contrast with other natural language tasks, and we investigate the reasons for this difference. We apply techniques that have been shown to improve performance on other NLP tasks and show that in most cases their impact on abstract reasoning performance is limited. In the course of this work, we have generated a new benchmark for evaluating language models on abstract reasoning tasks.

Neuromodulation Gated Transformer

We introduce a novel architecture, the Neuromodulation Gated Transformer (NGT), which is a simple implementation of neuromodulation in transformers via a multiplicative effect. We compare it to baselines and show that it results in the best average performance on the SuperGLUE benchmark validation sets.

Effectiveness of Debiasing Techniques: An Indigenous Qualitative Analysis

An indigenous perspective on the effectiveness of debiasing techniques for pre-trained language models (PLMs) is presented in this paper. The current techniques used to measure and debias PLMs are skewed towards the US racial biases and rely on pre-defined bias attributes (e.g. "black" vs "white"). Some require large datasets and further pre-training. Such techniques are not designed to capture the underrepresented indigenous populations in other countries, such as M\=aori in New Zealand. Local knowledge and understanding must be incorporated to ensure unbiased algorithms, especially when addressing a resource-restricted society.