arXiv subjects

Liam D. Turner

Publications and source records attributed to Liam D. Turner.

5 recordsLinked to original sources

arXiv2026

Shifting Perspectives: Steering Vectors for Robust Bias Mitigation in LLMs

We present a novel approach to bias mitigation in large language models (LLMs) by applying steering vectors to modify model activations in forward passes. We compute 8 steering vectors, each corresponding to a different social bias axis, such as age, gender, or race, on a training subset of the BBQ dataset and compare the effectiveness of these to 3 additional bias mitigation methods across 4 datasets. When optimized on the BBQ dataset, our individually tuned steering vectors achieve average improvements of 12.8% on BBQ, 8.3% on CLEAR-Bias, and 1% on StereoSet, and show improvements over prompting and Self-Debias in all cases, and improvements over fine-tuning in 12 out of 17 evaluations. In addition, steering vectors showed the lowest impact on MMLU scores of the four bias mitigation methods tested. The work presents the first systematic investigation of steering vectors for bias mitigation, and we demonstrate that they are a powerful and computationally efficient strategy for reducing bias in LLMs, with broader implications for enhancing AI safety.

cs.LG↗

arXiv2025

Dialz: A Python Toolkit for Steering Vectors

We introduce Dialz, a framework for advancing research on steering vectors for open-source LLMs, implemented in Python. Steering vectors allow users to modify activations at inference time to amplify or weaken a 'concept', e.g. honesty or positivity, providing a more powerful alternative to prompting or fine-tuning. Dialz supports a diverse set of tasks, including creating contrastive pair datasets, computing and applying steering vectors, and visualizations. Unlike existing libraries, Dialz emphasizes modularity and usability, enabling both rapid prototyping and in-depth analysis. We demonstrate how Dialz can be used to reduce harmful outputs such as stereotypes, while also providing insights into model behaviour across different layers. We release Dialz with full documentation, tutorials, and support for popular open-source models to encourage further research in safe and controllable language generation. Dialz enables faster research cycles and facilitates insights into model interpretability, paving the way for safer, more transparent, and more reliable AI systems.

cs.LG↗

arXiv2024

Who is better at math, Jenny or Jingzhen? Uncovering Stereotypes in Large Language Models

Large language models (LLMs) have been shown to propagate and amplify harmful stereotypes, particularly those that disproportionately affect marginalised communities. To understand the effect of these stereotypes more comprehensively, we introduce GlobalBias, a dataset of 876k sentences incorporating 40 distinct gender-by-ethnicity groups alongside descriptors typically used in bias literature, which enables us to study a broad set of stereotypes from around the world. We use GlobalBias to directly probe a suite of LMs via perplexity, which we use as a proxy to determine how certain stereotypes are represented in the model's internal representations. Following this, we generate character profiles based on given names and evaluate the prevalence of stereotypes in model outputs. We find that the demographic groups associated with various stereotypes remain consistent across model likelihoods and model outputs. Furthermore, larger models consistently display higher levels of stereotypical outputs, even when explicitly instructed not to.

cs.CL↗

arXiv2019

Understanding the Signature of Controversial Wikipedia Articles through Motifs in Editor Revision Networks

Wikipedia serves as a good example of how editors collaborate to form and maintain an article. The relationship between editors, derived from their sequence of editing activity, results in a directed network structure called the revision network, that potentially holds valuable insights into editing activity. In this paper we create revision networks to assess differences between controversial and non-controversial articles, as labelled by Wikipedia. Originating from complex networks, we apply motif analysis, which determines the under or over-representation of induced sub-structures, in this case triads of editors. We analyse 21,631 Wikipedia articles in this way, and use principal component analysis to consider the relationship between their motif subgraph ratio profiles. Results show that a small number of induced triads play an important role in characterising relationships between editors, with controversial articles having a tendency to cluster. This provides useful insight into editing behaviour and interaction capturing counter-narratives, without recourse to semantic analysis. It also provides a potentially useful feature for future prediction of controversial Wikipedia articles.

cs.SI↗

arXiv2018

Network Classification in Temporal Networks Using Motifs

Network classification has a variety of applications, such as detecting communities within networks and finding similarities between those representing different aspects of the real world. However, most existing work in this area focus on examining static undirected networks without considering directed edges or temporality. In this paper, we propose a new methodology that utilizes feature representation for network classification based on the temporal motif distribution of the network and a null model for comparing against random graphs. Experimental results show that our method improves accuracy by up $10\%$ compared to the state-of-the-art embedding method in network classification, for tasks such as classifying network type, identifying communities in email exchange network, and identifying users given their app-switching behaviors.

cs.SI↗