arXiv subjects

Mohamed Ahmed

Publications and source records attributed to Mohamed Ahmed.

14 recordsLinked to original sources

arXiv2026

Charged Higgs Pairs at the LHC: A NLO Analysis

Charged Higgs-boson pair production at hadron colliders yields information about the trilinear couplings involving charged Higgs fields in extensions of the Standard Model (SM). We consider the type-I two-Higgs doublet model (2HDM) extension and obtain next-to-leading order QCD predictions for the charged-Higgs pair production ($H^+H^-$ production). All production modes, i.e. Drell--Yan-like production, gluon fusion and vector-boson fusion are included in the analysis. We determine uncertainties originating from the scale dependence, the parton-density functions and strong coupling $α_s$ at the LHC. We observe that the QCD corrections lead to a significant reduction of the relative scale dependences. These improved predictions will allow for a quantitative interpretation of experimental measurements, in case that charged Higgs states will be observed.

hep-ph↗

arXiv2026

A Cherry-Picking Approach to Large Load Shaping for More Effective Carbon Reduction

Shaping multi-megawatt loads, such as data centers, impacts generator dispatch on the electric grid, which in turn affects system CO2 emissions and energy cost. Substantiating the effectiveness of prevalent load shaping strategies, such as those based on grid-level average carbon intensity, locational marginal price, or marginal emissions, is challenging due to the lack of detailed counterfactual data required for accurate attribution. This study uses a series of calibrated granular ERCOT day-ahead direct current optimal power flow (DC-OPF) simulations for counterfactual analysis of a broad set of load shaping strategies on grid CO2 emissions and cost of electricity. In terms of annual grid level CO2 emissions reductions, LMP-based shaping outperforms other common strategies, but can be significantly improved upon. Examining the performance of practicable strategies under different grid conditions motivates a more effective load shaping approach: one that "cherry-picks" a daily strategy based on observable grid signals and historical data. The cherry-picking approach to power load shaping is applicable to any large flexible consumer on the electricity grid, such as data centers, distributed energy resources and Virtual Power Plants (VPPs).

stat.ML↗

arXiv2025

Advancing Jailbreak Strategies: A Hybrid Approach to Exploiting LLM Vulnerabilities and Bypassing Modern Defenses

The advancement of Pre-Trained Language Models (PTLMs) and Large Language Models (LLMs) has led to their widespread adoption across diverse applications. Despite their success, these models remain vulnerable to attacks that exploit their inherent weaknesses to bypass safety measures. Two primary inference-phase threats are token-level and prompt-level jailbreaks. Token-level attacks embed adversarial sequences that transfer well to black-box models like GPT but leave detectable patterns and rely on gradient-based token optimization, whereas prompt-level attacks use semantically structured inputs to elicit harmful responses yet depend on iterative feedback that can be unreliable. To address the complementary limitations of these methods, we propose two hybrid approaches that integrate token- and prompt-level techniques to enhance jailbreak effectiveness across diverse PTLMs. GCG + PAIR and the newly explored GCG + WordGame hybrids were evaluated across multiple Vicuna and Llama models. GCG + PAIR consistently raised attack-success rates over its constituent techniques on undefended models; for instance, on Llama-3, its Attack Success Rate (ASR) reached 91.6%, a substantial increase from PAIR's 58.4% baseline. Meanwhile, GCG + WordGame matched the raw performance of WordGame maintaining a high ASR of over 80% even under stricter evaluators like Mistral-Sorry-Bench. Crucially, both hybrids retained transferability and reliably pierced advanced defenses such as Gradient Cuff and JBShield, which fully blocked single-mode attacks. These findings expose previously unreported vulnerabilities in current safety stacks, highlight trade-offs between raw success and defensive robustness, and underscore the need for holistic safeguards against adaptive adversaries.

cs.CL↗

arXiv2024

SummAct: Uncovering User Intentions Through Interactive Behaviour Summarisation

Recent work has highlighted the potential of modelling interactive behaviour analogously to natural language. We propose interactive behaviour summarisation as a novel computational task and demonstrate its usefulness for automatically uncovering latent user intentions while interacting with graphical user interfaces. To tackle this task, we introduce SummAct, a novel hierarchical method to summarise low-level input actions into high-level intentions. SummAct first identifies sub-goals from user actions using a large language model and in-context learning. High-level intentions are then obtained by fine-tuning the model using a novel UI element attention to preserve detailed context information embedded within UI elements during summarisation. Through a series of evaluations, we demonstrate that SummAct significantly outperforms baselines across desktop and mobile interfaces as well as interactive tasks by up to 21.9%. We further show three exciting interactive applications benefited from SummAct: interactive behaviour forecasting, automatic behaviour synonym identification, and language-based behaviour retrieval.

cs.HC↗

arXiv2024

METAL: Towards Multilingual Meta-Evaluation

With the rising human-like precision of Large Language Models (LLMs) in numerous tasks, their utilization in a variety of real-world applications is becoming more prevalent. Several studies have shown that LLMs excel on many standard NLP benchmarks. However, it is challenging to evaluate LLMs due to test dataset contamination and the limitations of traditional metrics. Since human evaluations are difficult to collect, there is a growing interest in the community to use LLMs themselves as reference-free evaluators for subjective metrics. However, past work has shown that LLM-based evaluators can exhibit bias and have poor alignment with human judgments. In this study, we propose a framework for an end-to-end assessment of LLMs as evaluators in multilingual scenarios. We create a carefully curated dataset, covering 10 languages containing native speaker judgments for the task of summarization. This dataset is created specifically to evaluate LLM-based evaluators, which we refer to as meta-evaluation (METAL). We compare the performance of LLM-based evaluators created using GPT-3.5-Turbo, GPT-4, and PaLM2. Our results indicate that LLM-based evaluators based on GPT-4 perform the best across languages, while GPT-3.5-Turbo performs poorly. Additionally, we perform an analysis of the reasoning provided by LLM-based evaluators and find that it often does not match the reasoning provided by human judges.

cs.CL↗

arXiv2024

Are Large Language Model-based Evaluators the Solution to Scaling Up Multilingual Evaluation?

Large Language Models (LLMs) excel in various Natural Language Processing (NLP) tasks, yet their evaluation, particularly in languages beyond the top $20$, remains inadequate due to existing benchmarks and metrics limitations. Employing LLMs as evaluators to rank or score other models' outputs emerges as a viable solution, addressing the constraints tied to human annotators and established benchmarks. In this study, we explore the potential of LLM-based evaluators, specifically GPT-4 in enhancing multilingual evaluation by calibrating them against $20$K human judgments across three text-generation tasks, five metrics, and eight languages. Our analysis reveals a bias in GPT4-based evaluators towards higher scores, underscoring the necessity of calibration with native speaker judgments, especially in low-resource and non-Latin script languages, to ensure accurate evaluation of LLM performance across diverse languages.

cs.CL↗

arXiv2022

A Few Thousand Translations Go a Long Way! Leveraging Pre-trained Models for African News Translation

Recent advances in the pre-training of language models leverage large-scale datasets to create multilingual models. However, low-resource languages are mostly left out in these datasets. This is primarily because many widely spoken languages are not well represented on the web and therefore excluded from the large-scale crawls used to create datasets. Furthermore, downstream users of these models are restricted to the selection of languages originally chosen for pre-training. This work investigates how to optimally leverage existing pre-trained models to create low-resource translation systems for 16 African languages. We focus on two questions: 1) How can pre-trained models be used for languages not included in the initial pre-training? and 2) How can the resulting translation models effectively transfer to new domains? To answer these questions, we create a new African news corpus covering 16 languages, of which eight languages are not part of any existing evaluation dataset. We demonstrate that the most effective strategy for transferring both to additional languages and to additional domains is to fine-tune large pre-trained models on small quantities of high-quality translation data.

cs.CL↗

arXiv2020

Molecular representation learning with language models and domain-relevant auxiliary tasks

We apply a Transformer architecture, specifically BERT, to learn flexible and high quality molecular representations for drug discovery problems. We study the impact of using different combinations of self-supervised tasks for pre-training, and present our results for the established Virtual Screening and QSAR benchmarks. We show that: i) The selection of appropriate self-supervised task(s) for pre-training has a significant impact on performance in subsequent downstream tasks such as Virtual Screening. ii) Using auxiliary tasks with more domain relevance for Chemistry, such as learning to predict calculated molecular properties, increases the fidelity of our learnt representations. iii) Finally, we show that molecular representations learnt by our model `MolBert' improve upon the current state of the art on the benchmark datasets.

cs.LG↗

arXiv2018

DEFactor: Differentiable Edge Factorization-based Probabilistic Graph Generation

Generating novel molecules with optimal properties is a crucial step in many industries such as drug discovery. Recently, deep generative models have shown a promising way of performing de-novo molecular design. Although graph generative models are currently available they either have a graph size dependency in their number of parameters, limiting their use to only very small graphs or are formulated as a sequence of discrete actions needed to construct a graph, making the output graph non-differentiable w.r.t the model parameters, therefore preventing them to be used in scenarios such as conditional graph generation. In this work we propose a model for conditional graph generation that is computationally efficient and enables direct optimisation of the graph. We demonstrate favourable performance of our model on prototype-based molecular graph conditional generation tasks.

cs.LG↗

arXiv2016

Learning Convolutional Neural Networks for Graphs

Numerous important problems can be framed as learning from graph data. We propose a framework for learning convolutional neural networks for arbitrary graphs. These graphs may be undirected, directed, and with both discrete and continuous node and edge attributes. Analogous to image-based convolutional networks that operate on locally connected regions of the input, we present a general approach to extracting locally connected regions from graphs. Using established benchmark data sets, we demonstrate that the learned feature representations are competitive with state of the art graph kernels and that their computation is highly efficient.

cs.LG↗

arXiv2016

Understanding the Detection of View Fraud in Video Content Portals

While substantial effort has been devoted to understand fraudulent activity in traditional online advertising (search and banner), more recent forms such as video ads have received little attention. The understanding and identification of fraudulent activity (i.e., fake views) in video ads for advertisers, is complicated as they rely exclusively on the detection mechanisms deployed by video hosting portals. In this context, the development of independent tools able to monitor and audit the fidelity of these systems are missing today and needed by both industry and regulators. In this paper we present a first set of tools to serve this purpose. Using our tools, we evaluate the performance of the audit systems of five major online video portals. Our results reveal that YouTube's detection system significantly outperforms all the others. Despite this, a systematic evaluation indicates that it may still be susceptible to simple attacks. Furthermore, we find that YouTube penalizes its videos' public and monetized view counters differently, the former being more aggressive. This means that views identified as fake and discounted from the public view counter are still monetized. We speculate that even though YouTube's policy puts in lots of effort to compensate users after an attack is discovered, this practice places the burden of the risk on the advertisers, who pay to get their ads displayed.

cs.CY↗

arXiv2015

Unravelling the Impact of Temporal and Geographical Locality in Content Caching Systems

To assess the performance of caching systems, the definition of a proper process describing the content requests generated by users is required. Starting from the analysis of traces of YouTube video requests collected inside operational networks, we identify the characteristics of real traffic that need to be represented and those that instead can be safely neglected. Based on our observations, we introduce a simple, parsimonious traffic model, named Shot Noise Model (SNM), that allows us to capture temporal and geographical locality of content popularity. The SNM is sufficiently simple to be effectively employed in both analytical and scalable simulative studies of caching systems. We demonstrate this by analytically characterizing the performance of the LRU caching policy under the SNM, for both a single cache and a network of caches. With respect to the standard Independent Reference Model (IRM), some paradigmatic shifts, concerning the impact of various traffic characteristics on cache performance, clearly emerge from our results.

cs.NI↗

arXiv2013

Temporal Locality in Today's Content Caching: Why it Matters and How to Model it

The dimensioning of caching systems represents a difficult task in the design of infrastructures for content distribution in the current Internet. This paper addresses the problem of defining a realistic arrival process for the content requests generated by users, due its critical importance for both analytical and simulative evaluations of the performance of caching systems. First, with the aid of YouTube traces collected inside operational residential networks, we identify the characteristics of real traffic that need to be considered or can be safely neglected in order to accurately predict the performance of a cache. Second, we propose a new parsimonious traffic model, named the Shot Noise Model (SNM), that enables users to natively capture the dynamics of content popularity, whilst still being sufficiently simple to be employed effectively for both analytical and scalable simulative studies of caching systems. Finally, our results show that the SNM presents a much better solution to account for the temporal locality observed in real traffic compared to existing approaches.

cs.NI↗

arXiv2013

Analyzing the Performance of LRU Caches under Non-Stationary Traffic Patterns

This work presents, to the best of our knowledge of the literature, the first analytic model to address the performance of an LRU (Least Recently Used) implementing cache under non-stationary traffic conditions, i.e., when the popularity of content evolves with time. We validate the accuracy of the model using Monte Carlo simulations. We show that the model is capable of accurately estimating the cache hit probability, when the popularity of content is non-stationary. We find that there exists a dependency between the performance of an LRU implementing cache and i) the lifetime of content in a system, ii) the volume of requests associated with it, iii) the distribution of content request volumes and iv) the shape of the popularity profile over time.

cs.NI↗