publications

2026

1. arXiv

   Training and Agentic Inference Strategies for LLM-based Manim Animation Generation

   Ravidu Suien Rammuni Silva, Ahmad Lotfi, Isibor Kennedy Ihianle, and 2 more authors

   arXiv preprint arXiv:2604.18364, 2026

   Abs DOI arXiv Bib

   Generating programmatic animation using libraries such as Manim presents unique challenges for Large Language Models (LLMs), requiring spatial reasoning, temporal sequencing, and familiarity with domain-specific APIs that are underrepresented in general pre-training data. A systematic study of how training and inference strategies interact in this setting is lacking in current research. This study introduces ManimTrainer, a training pipeline that combines Supervised Fine-tuning (SFT) with Reinforcement Learning (RL) based Group Relative Policy Optimisation (GRPO) using a unified reward signal that fuses code and visual assessment signals, and ManimAgent, an inference pipeline featuring Renderer-in-the-loop (RITL) and API documentation-augmented RITL (RITL-DOC) strategies. Using these techniques, this study presents the first unified training and inference study for text-to-code-to-video transformation with Manim. It evaluates 17 open-source sub-30B LLMs across nine combinations of training and inference strategies using ManimBench. Results show that SFT generally improves code quality, while GRPO enhances visual outputs and increases the models’ responsiveness to extrinsic signals during self-correction at inference time. The Qwen 3 Coder 30B model with GRPO and RITL-DOC achieved the highest overall performance, with a 94% Render Success Rate (RSR) and 85.7% Visual Similarity (VS) to reference videos, surpassing the baseline GPT-4.1 model by +3 percentage points in VS. Additionally, the analysis shows that the correlation between code and visual metrics strengthens with SFT and GRPO but weakens with inference-time enhancements, highlighting the complementary roles of training and agentic inference strategies in Manim animation generation.

   @article{silva2026training,
     title = {Training and Agentic Inference Strategies for LLM-based Manim Animation Generation},
     author = {Rammuni Silva, Ravidu Suien and Lotfi, Ahmad and Ihianle, Isibor Kennedy and Shahtahmassebi, Golnaz and Bird, Jordan J},
     journal = {arXiv preprint arXiv:2604.18364},
     year = {2026},
     doi = {10.48550/arXiv.2604.18364}
   }

2. arXiv

   LessonBench-V1: A Benchmark Dataset for Evaluating AI Lesson Generation Agents

   Ravidu Suien Rammuni Silva, Ahmad Lotfi, Isibor Kennedy Ihianle, and 2 more authors

   arXiv preprint arXiv:2607.13041, 2026

   Abs DOI arXiv Bib

   Large Language Model (LLM) based AI educational content generation systems are increasingly being developed, yet no standardised benchmark exists to systematically evaluate them. This study introduces LessonBench-V1, a benchmark dataset comprising 647 human-written lessons paired with LLM-based reverse-engineered lesson plans across 240 STEM topics spanning mathematics, physics, chemistry, and computer science. The lessons are drawn from 97 trusted open sources, including LibreTexts, Brilliant.org and GeeksForGeeks. Each lesson plan is human-reviewed and produced through a pedagogically grounded methodology that synthesises Bloom’s Taxonomy, Gagné’s Events, Merrill’s First Principles, and the 5E Instructional Model. The lesson plans capture 3,620 learning objectives with pedagogical metadata, enabling systematic, reproducible evaluation of lesson-generation AI agents and supporting further research. The study further proposes a three-dimensional evaluation pipeline for use with the dataset.

   @article{silva2026lessonbench,
     title = {LessonBench-V1: A Benchmark Dataset for Evaluating AI Lesson Generation Agents},
     author = {Rammuni Silva, Ravidu Suien and Lotfi, Ahmad and Ihianle, Isibor Kennedy and Shahtahmassebi, Golnaz and Bird, Jordan J},
     journal = {arXiv preprint arXiv:2607.13041},
     year = {2026},
     doi = {10.48550/arXiv.2607.13041}
   }

3. UKCI

   Large Language Model Approaches to Educational Video Generation Using Manim

   Ravidu Suien Rammuni Silva, Ahmad Lotfi, Isibor Kennedy Ihianle, and 2 more authors

   In Advances in Computational Intelligence Systems (UKCI 2025), 2026

   Abs DOI Bib

   The automated generation of educational videos based on Large Language Models (LLMs) holds promising potential to enhance pedagogy. This study investigates the use of compact LLMs with fewer than 10 billion parameters, utilising Parameter-Efficient Fine-Tuning (PEFT) techniques to generate Manim Python code that produces images and videos for educational visualisations from natural language narratives. This approach is highly resource-efficient compared to diffusion-based video generation, which is often infeasible on consumer-level hardware. Due to the lack of data availability, this study curated and reviewed the first benchmarking dataset for Manim code generation, ManimBench. ManimBench features 417 paired descriptions and corresponding Manim code samples categorised as basic, intermediate, and advanced, sourced directly from the official ManimCE documentation. The curated dataset was then used to fine-tune seven open-source compact LLMs using LoRA and QLoRA. Subsequently, their performances were evaluated comparatively against their base models using both functional and semantic metrics. The results indicate that the LoRA Fine-Tuned (FT) models consistently outperform the base models, achieving a 62% Manim render success rate. By leveraging ManimBench, further comparisons were made between the FT models and larger 14B parameter LLMs, finding that the PEFT-tuned FT coder models perform comparably well in accurate Manim code generation. However, it was also discovered that LLMs frequently make hallucination-related API errors in the generated code.

   @inproceedings{silva1468large,
     title = {Large Language Model Approaches to Educational Video Generation Using Manim},
     author = {Rammuni Silva, Ravidu Suien and Lotfi, Ahmad and Ihianle, Isibor Kennedy and Shahtahmassebi, Golnaz and Bird, Jordan J},
     booktitle = {Advances in Computational Intelligence Systems (UKCI 2025)},
     series = {Advances in Intelligent Systems and Computing},
     volume = {1468},
     pages = {306--317},
     year = {2026},
     publisher = {Springer Nature Switzerland},
     doi = {10.1007/978-3-032-07938-1_26}
   }

4. NeurIPS-W

   Visually Grounded Reinforcement Learning for LLM Programmatic Animation Generation in Interactive Execution Environments

   Ravidu Suien Rammuni Silva, Ahmad Lotfi, Isibor Kennedy Ihianle, and 2 more authors

   In NeurIPS 2026 Workshop on Foundations of LLM Post-Training in Changing Environments, Dec 2026

   Abs Bib PDF Code

   Generating programmatic animations with libraries such as Manim poses unique challenges for large language models (LLMs), requiring spatial reasoning, temporal sequencing, and adaptation to interactive execution environments. However, how different post-training objectives influence an LLM’s capacity to leverage dynamic feedback during generation remains under-researched. This study introduces ManimTrainer, a framework that combines Supervised Fine-tuning (SFT) with visually grounded Reinforcement Learning via Group Relative Policy Optimisation (GRPO) using a unified code-and-visual reward signal, alongside ManimAgent, an inference pipeline featuring Renderer-in-the-loop (RITL) and documentation-augmented execution feedback. Evaluating 17 open-source sub-30B LLMs across nine combinations of training and inference strategies on ManimBench, this work presents a unified study investigating how training types interact with test-time execution feedback. The results show that SFT and GRPO improve visual quality to a comparable degree under vanilla inference, while GRPO also preserves the policy’s responsiveness to extrinsic error signals during self-correction. Under renderer feedback, GRPO outperforms the base and SFT models by median margins of 2.7 pp and 1.3 pp across the 17 models, and the same trend also holds under RITL-DOC. Additionally, the analysis shows that the correlation between code and visual metrics strengthens with SFT and GRPO but weakens under inference-time execution loops, highlighting the complementary roles of training objectives and interactive environments in programmatic generation.

   @inproceedings{rammunisilva2026visually,
     title = {Visually Grounded Reinforcement Learning for {LLM} Programmatic Animation Generation in Interactive Execution Environments},
     author = {Rammuni Silva, Ravidu Suien and Lotfi, Ahmad and Ihianle, Isibor Kennedy and Shahtahmassebi, Golnaz and Bird, Jordan J.},
     booktitle = {NeurIPS 2026 Workshop on Foundations of LLM Post-Training in Changing Environments},
     year = {2026},
     month = dec,
     url = {https://ravidusilva.com/assets/pdf/rammunisilva2026visually.pdf},
   }

2025

1. Preprint

   Generative Artificial Intelligence and the Future of Education: A Systematic Review of Trends, Trustworthiness and Technological Evaluation

   Ravidu Suien Rammuni Silva, Ahmad Lotfi, Isibor Kennedy Ihianle, and 2 more authors

   Research Square preprint, 2025

   Abs DOI Bib

   This systematic review aims to investigate the current landscape of AI in Education (AIEd) through a comprehensive analysis of empirical studies related to AIEd published since 2022 and indexed in the Web of Science. Following the PRISMA 2020 guidelines and using thematic analysis, this study reviews 103 studies selected by one primary researcher based on the defined research questions, with four researchers independently verifying the selection. These research questions focus on the prevalent AIEd technologies, their reliability, evaluation frameworks, and the potential impact of AI in educational contexts, aiming to serve as a roadmap for researchers, developers, educators, and policymakers navigating the rapidly transforming landscape of AIEd. The findings highlight the prominence of classical machine learning models in published studies, while recent Generative AI (GenAI) models, such as Large Language Models (LLMs), have not been comprehensively experimented with, despite their potential. Findings show that studies utilising LLMs typically relied on more general-purpose systems, such as ChatGPT. Recognising these trends and analysing surveys on stakeholder intentions (educators and learners), the review identifies existing gaps in AIEd research, including the absence of adequate evaluation mechanisms, the lack of specialised LLM-based systems, and notable concerns about transparency, trustworthiness, and ethical implications. Based on the identified gaps, predominantly from a technical perspective, the study recommends research on specialised GenAI-based systems, the integration of AI tools into current pedagogical approaches, comprehensive ethics-aware evaluation frameworks, and addressing issues regarding the trustworthiness of AIEd systems.

   @article{silva2025generative,
     title = {Generative Artificial Intelligence and the Future of Education: A Systematic Review of Trends, Trustworthiness and Technological Evaluation},
     author = {Rammuni Silva, Ravidu Suien and Lotfi, Ahmad and Ihianle, Isibor Kennedy and Shahtahmassebi, Golnaz and Bird, Jordan J},
     journal = {Research Square preprint},
     year = {2025},
     doi = {10.21203/rs.3.rs-8248369/v1}
   }

2024

1. arXiv

   ArtBrain: An Explainable end-to-end Toolkit for Classification and Attribution of AI-Generated Art and Style

   Ravidu Suien Rammuni Silva, Ahmad Lotfi, Isibor Kennedy Ihianle, and 2 more authors

   arXiv preprint arXiv:2412.01512, 2024

   Abs DOI arXiv Bib Code

   Recently, the quality of artworks generated using Artificial Intelligence (AI) has increased significantly, resulting in growing difficulties in detecting synthetic artworks. However, limited studies have been conducted on identifying the authenticity of synthetic artworks and their source. This paper introduces AI-ArtBench, a dataset featuring 185,015 artistic images across 10 art styles. It includes 125,015 AI-generated images and 60,000 pieces of human-created artwork. This paper also outlines a method to accurately detect AI-generated images and trace them to their source model. This work proposes a novel Convolutional Neural Network model based on the ConvNeXt model called AttentionConvNeXt. AttentionConvNeXt was implemented and trained to differentiate between the source of the artwork and its style with an F1-Score of 0.869. The accuracy of attribution to the generative model reaches 0.999. To combine the scientific contributions arising from this study, a web-based application named ArtBrain was developed to enable both technical and non-technical users to interact with the model. Finally, this study presents the results of an Artistic Turing Test conducted with 50 participants. The findings reveal that humans could identify AI-generated images with an accuracy of approximately 58%, while the model itself achieved a significantly higher accuracy of around 99%.

   @article{silva2024artbrain,
     title = {ArtBrain: An Explainable end-to-end Toolkit for Classification and Attribution of AI-Generated Art and Style},
     author = {Rammuni Silva, Ravidu Suien and Lotfi, Ahmad and Ihianle, Isibor Kennedy and Shahtahmassebi, Golnaz and Bird, Jordan J},
     journal = {arXiv preprint arXiv:2412.01512},
     year = {2024},
     doi = {10.48550/arXiv.2412.01512}
   }

2023

1. arXiv

   FM-G-CAM: A Holistic Approach for Explainable AI in Computer Vision

   Ravidu Suien Rammuni Silva and Jordan J Bird

   arXiv preprint arXiv:2312.05975, 2023

   Abs DOI arXiv Bib Code

   Explainability is a vital aspect of modern AI for real-world impact and usability. The main objective of this paper is to emphasise the need to understand the predictions of Computer Vision models, specifically Convolutional Neural Network (CNN) models. Existing methods for explaining CNN predictions are largely based on Gradient-weighted Class Activation Maps (Grad-CAM) and focus solely on a single target class; this assumption about the target class selection neglects a large portion of the predictor CNN’s prediction process. In this paper, we present an exhaustive methodology, called Fused Multi-class Gradient-weighted Class Activation Map (FM-G-CAM), that considers multiple top-predicted classes and provides a holistic explanation of the predictor CNN’s rationale. We also provide a detailed mathematical and algorithmic description of our method. Furthermore, alongside a concise comparison of existing methods, we compare FM-G-CAM with Grad-CAM, quantitatively and qualitatively highlighting its benefits through real-world practical use cases. Finally, we present an open-source Python library with an FM-G-CAM implementation to conveniently generate saliency maps for CNN-based model predictions.

   @article{silva2023fm,
     title = {FM-G-CAM: A Holistic Approach for Explainable AI in Computer Vision},
     author = {Rammuni Silva, Ravidu Suien and Bird, Jordan J},
     journal = {arXiv preprint arXiv:2312.05975},
     year = {2023},
     doi = {10.48550/arXiv.2312.05975}
   }

2022

1. SN Comp. Sci.

   Effective Utilization of Multiple Convolutional Neural Networks for Chest X-Ray Classification

   Ravidu Suien Rammuni Silva and Pumudu Fernando

   SN Computer Science, 2022

   Abs DOI Bib Code

   Out of the numerous types of Medical Imaging modalities available, radiography stands out a bit more than others due to its capabilities of diagnosing diseases and conditions, including life-threatening conditions. Its affordability is another main reason for its prevalence. Chest Radiography holds even higher importance, as it focuses a critical area of the human body. However, interpreting a Chest Radiography image can be challenging and usually done by an experienced Radiologist for accurate results. There are two main issues related to this. One is that in some countries, experienced Radiologists are scarce. The other issue is that the inevitability of human errors in diagnoses. Researchers attempt to use Artificial Intelligence to address these two issues. Most of the existing work incorporates Convolutional Neural Networks for this purpose. This paper presents a novel way of parallelizing multiple architectures of Convolutional Neural Networks focusing on Chest X-ray classification. The paper further presents a comprehensive evaluation of the existing architectures with the parallelized results of them using our method. We used four large-scale datasets, including a non-medical one, for the evaluation of our models. We managed to achieve better accuracy for 9 out 13 and 11 out of 14 labels on our two main evaluation datasets. The paper concludes by presenting the limitations and future improvements possible for the system.

   @article{rammuni2022effective,
     title = {Effective Utilization of Multiple Convolutional Neural Networks for Chest X-Ray Classification},
     author = {Rammuni Silva, Ravidu Suien and Fernando, Pumudu},
     journal = {SN Computer Science},
     volume = {3},
     number = {6},
     pages = {492},
     year = {2022},
     publisher = {Springer Nature Singapore},
     doi = {10.1007/s42979-022-01390-9}
   }

2. NILES

   Recent Developments and Limitations in Recommender Systems: A Review

   Tharinda Dilshan Piyadasa, Ravidu Suien Rammuni Silva, and Dharshana Kasthurirathna

   In 2022 4th Novel Intelligent and Leading Emerging Sciences Conference (NILES), 2022

   Abs DOI Bib

   Recommender systems are an indispensable and inseparable part of everyday online content offerings like music, movies, sports, and e-commerce. They help users by personalizing products, services, and content, cutting the time and effort required to browse through large amounts of online information. Despite the evolution of recommender systems, the extent to which they can be explored and extended is yet to be discovered. This paper conducts an extensive review of recent developments in recommender systems and presents a comprehensive analysis, highlighting their approaches and limitations. Based on the findings of the analysis, recommendations are provided on possible research avenues that look promising to investigate further. The primary objective of this study is to provide an entry point for researchers who intend to address the current limitations of recommender systems and develop new strategies. Based on the recent breakthroughs in recommender systems detailed in this study, it is possible to find various prospective research avenues, not limited to novel recommender systems but also new evaluation mechanisms and security measures.

   @inproceedings{piyadasa2022recent,
     title = {Recent Developments and Limitations in Recommender Systems: A Review},
     author = {Piyadasa, Tharinda Dilshan and Rammuni Silva, Ravidu Suien and Kasthurirathna, Dharshana},
     booktitle = {2022 4th Novel Intelligent and Leading Emerging Sciences Conference (NILES)},
     pages = {81--85},
     year = {2022},
     organization = {IEEE},
     doi = {10.1109/NILES56402.2022.9942405}
   }

2021

1. FRUCT

   An Extensive Survey of Machine Learning Based Approaches on Automated Pathology Detection in Chest X-Rays

   Ravidu Suien Rammuni Silva and Pumudu Fernando

   In 2021 28th Conference of Open Innovations Association (FRUCT), 2021

   Abs DOI Bib

   Radiography is one of the most common and eminent medical imaging technologies in the world to date. Chest radiography is a very powerful and successful way of diagnosing thoracic diseases of humans. With the latest advancements and development in computer hardware, computer vision and especially with the publicly available large-scale datasets, machine learning based approaches on automated pathology detection in chest radiography have become increasingly popular among researchers. Our study conducts an extensive survey on existing machine learning approaches, its datasets and techniques on pathology detection in Chest X-Rays. The paper presents popular and publicly available labelled Chest X-Rays datasets with its specifications and discusses about the labellers, labelling methodologies used by them in a comprehensive discussion. Then, popular effective Image Processing techniques for Chest X-Rays images are presented. Then the paper further discusses about the current machine learning architectures used and portraits the effectiveness of Deep Convolutional Neural Networks for the purpose. Finally, the paper concludes with a discussion with gaps in current literature, unexplored areas and possible future with them in Machine Learning based automated pathology detection on Chest X-Rays.

   @inproceedings{silva2021extensive,
     title = {An Extensive Survey of Machine Learning Based Approaches on Automated Pathology Detection in Chest X-Rays},
     author = {Rammuni Silva, Ravidu Suien and Fernando, Pumudu},
     booktitle = {2021 28th Conference of Open Innovations Association (FRUCT)},
     pages = {365--373},
     year = {2021},
     organization = {IEEE},
     doi = {10.23919/FRUCT50888.2021.9347605}
   }