School of Computing Science

[FATA Seminar] Efficient Type-checking of a Dependently Typed Programming Language

Group: Formal Analysis, Theory and Algorithms (FATA)
Speaker: Edwin Brady, University of St Andrews
Date: 07 April, 2026
Time: 15:00 - 16:00
Location: Room 422, SAWB

Idris is a dependently typed programming language, developed in St Andrews with many contributions from elsewhere, used for research in programming language design and verified software development. While the compiler is, in most everyday cases, efficient enough (for example it is implemented in itself and compiles itself and its libraries within a few minutes) we have found that for a certain type of program, type-checking can be a significant bottleneck. In this talk I will briefly describe how a type-checker for a dependently typed language works, give a small example which illustrates the problem, and give an overview of a new design for the type-checker currently being implemented in a new version of Idris. I will also, time permitting, discuss lessons we have learned about dependently typed programming language design which affect performance and usability, and which will influence the next version.

Multivector Reranking in the Era of Strong First-Stage Retrievers

Group: Information Retrieval (IR)
Speaker: Silvio Martinico, University of Pisa
Date: 07 April, 2026
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 423 Seminar Room

Title
Multivector Reranking in the Era of Strong First-Stage Retrievers

Abstract
Multivector late-interaction models have dramatically improved retrieval effectiveness by enabling fine-grained token-level interactions, but their high computational cost severely limits real-world deployment. To mitigate this, most modern systems rely on a gather-and-refine strategy, where individual token vectors gather candidate documents that are then fully scored. However, this approach still requires expensive searches over large token sets and often drops highly relevant documents before they can be scored.
In this talk, we will explore how to address this bottleneck by recasting the pipeline into a highly optimized, two-stage retrieval architecture. By replacing the token-level gather phase with a learned sparse retriever (LSR), we can generate a smaller, more semantically coherent candidate set and drastically reduce both gather time and the number of candidates required for full rescoring. We also discuss practical optimization techniques that tune the pipeline and adapt it to each query. Finally, we describe how tools such as kANNolo, Vectorium, and Seismic support the development of this architecture, and outline open challenges and emerging directions for efficient multivector retrieval.

Bio
Silvio Martinico is a Ph.D. student in Computer Science at the University of Pisa, supervised by Franco Maria Nardini, Cosimo Rulli, and Rossano Venturini. His research focuses on Information Retrieval, specifically targeting the algorithmic efficiency of Approximate Nearest Neighbors (ANN) search and large-scale multivector retrieval. His work aims to overcome critical computational bottlenecks to make highly expressive retrieval models feasible at scale. He is also the creator of kANNolo, a research-oriented Rust library for ANN search.

CyberUK Workshop Dry Run

Group: Networked Systems Research Laboratory (NETLAB)
Speaker: Marco Cook
Date: 09 April, 2026
Time: 11:00 - 12:00
Location: Lilybank Gardens, F121 Conference Room

We have an upcoming workshop at CyberUK, and we’d like to run a dry session with the group to ensure the activities are clear and can be completed within the allocated time.

[FATA Seminar] Efficient Type-checking of a Dependently Typed Programming Language

Group: Formal Analysis, Theory and Algorithms (FATA)
Speaker: Edwin Brady, University of St Andrews
Date: 07 April, 2026
Time: 15:00 - 16:00
Location: Room 422, SAWB

Idris is a dependently typed programming language, developed in St Andrews with many contributions from elsewhere, used for research in programming language design and verified software development. While the compiler is, in most everyday cases, efficient enough (for example it is implemented in itself and compiles itself and its libraries within a few minutes) we have found that for a certain type of program, type-checking can be a significant bottleneck. In this talk I will briefly describe how a type-checker for a dependently typed language works, give a small example which illustrates the problem, and give an overview of a new design for the type-checker currently being implemented in a new version of Idris. I will also, time permitting, discuss lessons we have learned about dependently typed programming language design which affect performance and usability, and which will influence the next version.

Multivector Reranking in the Era of Strong First-Stage Retrievers

Group: Information Retrieval (IR)
Speaker: Silvio Martinico, University of Pisa
Date: 07 April, 2026
Time: 15:00 - 16:00
Location: Sir Alwyn Williams Building, 423 Seminar Room

Title
Multivector Reranking in the Era of Strong First-Stage Retrievers

Abstract
Multivector late-interaction models have dramatically improved retrieval effectiveness by enabling fine-grained token-level interactions, but their high computational cost severely limits real-world deployment. To mitigate this, most modern systems rely on a gather-and-refine strategy, where individual token vectors gather candidate documents that are then fully scored. However, this approach still requires expensive searches over large token sets and often drops highly relevant documents before they can be scored.
In this talk, we will explore how to address this bottleneck by recasting the pipeline into a highly optimized, two-stage retrieval architecture. By replacing the token-level gather phase with a learned sparse retriever (LSR), we can generate a smaller, more semantically coherent candidate set and drastically reduce both gather time and the number of candidates required for full rescoring. We also discuss practical optimization techniques that tune the pipeline and adapt it to each query. Finally, we describe how tools such as kANNolo, Vectorium, and Seismic support the development of this architecture, and outline open challenges and emerging directions for efficient multivector retrieval.

Bio
Silvio Martinico is a Ph.D. student in Computer Science at the University of Pisa, supervised by Franco Maria Nardini, Cosimo Rulli, and Rossano Venturini. His research focuses on Information Retrieval, specifically targeting the algorithmic efficiency of Approximate Nearest Neighbors (ANN) search and large-scale multivector retrieval. His work aims to overcome critical computational bottlenecks to make highly expressive retrieval models feasible at scale. He is also the creator of kANNolo, a research-oriented Rust library for ANN search.

CyberUK Workshop Dry Run

Group: Networked Systems Research Laboratory (NETLAB)
Speaker: Marco Cook
Date: 09 April, 2026
Time: 11:00 - 12:00
Location: Lilybank Gardens, F121 Conference Room

We have an upcoming workshop at CyberUK, and we’d like to run a dry session with the group to ensure the activities are clear and can be completed within the allocated time.

Upwards x GWiCS Seminar: Academia or Industry – Post-PhD Career Paths of Two Recent SoCS Graduates

Group: School of Computing Science
Speaker: Dr Jacqueline Borgstedt / Dr Robert Szafarczyk, ETH Zurich / Qualcomm
Date: 13 April, 2026
Time: 12:00 - 13:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

Topic: Academia or Industry – Post-PhD Career Paths of Two Recent SoCS Graduates

Speakers:

- Dr Jacqueline Borgstedt
(Postdoc at ETH Zurich, https://sbs.ethz.ch/ people/postdocs/jacqueline-borgstedt.html)

- Dr Robert Szafarczyk
(Senior Engineer at Qualcomm, https:// robertszafa.github.io/)


Location:

- in Room SAWB 422

- and on Zoom: https://uofglasgow.zoom.us/j/88322102253? pwd=VyI6hGJosem9FzUZbYMSwHxmoo6fY4.1


This is a joint Upwards and GWiCS seminar.

As always, everyone in SoCS is welcome, but this session is aimed mainly at PhD students and postdocs.

 
What will this session be about?
 
It is up to the speakers to set the agenda for their Upwards/GWiCS talks, but the idea of this seminar instance is to hear what two SoCS graduates have been up to since completing their PhDs in our School recently.

Some of the questions we suggested for the two talks are: What were the different options you considered after your PhD? Why did you decide to go on in academia/industry? Why for a postdoc at ETH / engineering role at Qualcomm? How is your role now different from being a PhD student in the School of Computing Science at Glasgow? What did you do in your PhD project, and what do you do now in your project as a postdoc/engineer? What are your longer-term career aspirations?

What is Upwards?
 
Upwards is the School of Computing Science’s research culture seminar, covering all facets of developing, conducting, and disseminating research and related topics (e.g. managing a research team, time management to do research, connections between research and teaching). It is open to everyone in the School, but a specific aim is to support ECR development and some sessions are aimed mainly at PGRs and/or PDRAs.

How are the seminars held?

Upwards seminars are held in person in the School to bring people together. In addition, the sessions are streamed on Zoom to allow to join remotely, if attending in person is not an option. To preserve the off-the- record atmosphere of the seminars, which allows the speakers to speak more freely about their personal experiences, the seminars are not recorded and the slides are not shared. For the same reason, AI tools (such as those that automatically take meeting notes) will not be permitted. 

tbc

Group: Systems Seminars
Speaker: Fangzhou "Bonnie" Dong, Arizona State University
Date: 14 April, 2026
Time: 16:00 - 17:00
Location: Room 422, Sir Alwyn Williams Building and Zoom

tbc

[FATA Seminar] TBA

Group: Formal Analysis, Theory and Algorithms (FATA)
Speaker: Wietze Koops, Lund University
Date: 16 April, 2026
Time: 11:00 - 12:00
Location: Lilybank Gardens, F121 Conference Room

TBA

SICSA Writing Retreat 2026

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 27 April, 2026
Time: 15:00 - 14:00
Location: Scottish Centre for Ecology and the Natural Environment (SCENE), G63 0JS

The 2026 SICSA Writing Retreat will bring together researchers from across Scotland for a two-day intensive writing event. The programme will consist of networking and skill sharing activities, in addition to individual and group writing blocks. Postdoctoral researchers from any SICSA institution are invited to apply to attend the writing retreat by completing the online form by 1 February 2026. Spaces are very limited and the SICSA Directorate will be judging applications based on clear and achievable writing plans, quality outputs and benefits to both individual researchers and wider groups. Proposals that involve and benefit multiple SICSA institutions are particularly encouraged. Apply Date Start: 15:00 Monday 27 April 2026 Finish: 14:00 Wednesday 29 April 2026 Location Scottish Centre for Ecology and the Natural Environment (SCENE)

SICSA Writing Retreat 2026

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 27 April, 2026
Time: 15:00 - 14:00
Location: Scottish Centre for Ecology and the Natural Environment (SCENE), G63 0JS

The 2026 SICSA Writing Retreat will bring together researchers from across Scotland for a two-day intensive writing event. The programme will consist of networking and skill sharing activities, in addition to individual and group writing blocks. Postdoctoral researchers from any SICSA institution are invited to apply to attend the writing retreat by completing the online form by 1 February 2026. Spaces are very limited and the SICSA Directorate will be judging applications based on clear and achievable writing plans, quality outputs and benefits to both individual researchers and wider groups. Proposals that involve and benefit multiple SICSA institutions are particularly encouraged. Apply Date Start: 15:00 Monday 27 April 2026 Finish: 14:00 Wednesday 29 April 2026 Location Scottish Centre for Ecology and the Natural Environment (SCENE)

Synergistic Hardware-Software Co-Design for Approximate Real-Time Systems

Group: Systems Seminars
Speaker: Shounak Chakraborty, Durham University
Date: 28 April, 2026
Time: 14:00 - 15:00
Location: Room 422, Sir Alwyn Williams Building and Zoom

Abstract: Modern data centers in hubs consume hundreds of megawatts to support global services. To address this energy crisis, we must optimize the performance-per-watt across the entire stack—from application layers down to the device level. This seminar introduces PRECIOUS, a framework designed to maximize Quality of Service (QoS) for dependent real-time tasks on heterogeneous Chip Multi-Processors (CMPs). The core of PRECIOUS is a hybrid offline-online strategy. Offline, we utilise an Integer Linear Programming (ILP) scheduling approach to optimally assign task versions and cores while satisfying power and timing constraints. Online, the framework leverages the density of Multi-Level Cell (MLC) MRAM-based Last-Level Caches (LLC). By employing novel cell-splitting and intelligent write-steering, we reduce cache miss rates by 19% and improve throughput by 5.7%. Validated on 64-core systems, PRECIOUS achieves up to 76% normalised QoS, outperforming traditional heuristic methods. Furthermore, the framework converts architectural efficiency into dynamic runtime slacks, enabling a 9.0% QoS boost and cluster power-gating without additional energy overhead. 

Speaker's Bio: Dr. Shounak Chakraborty is an Assistant Professor in the Department of Computer Science at Durham University and a member of the Scientific Computing Research Group. He also serves as a Visiting Fellow at the University of Essex and was previously a Marie Curie Postdoctoral Fellow as well as an ERCIM Postdoctoral Fellow at NTNU, Norway. Beyond his academic background, he brings two years of industry experience as a Computer System Architect at ZeroPoint Technologies AB in Sweden, where he researched memory compression mechanisms to enhance energy efficiency. His research is situated at the intersection of computer architecture and compilers, with a focus on improving the energy and thermal efficiency of modern Chip Multi-Processors. Dr. Chakraborty's recent work, which includes a project supported by APRIL AI Hub, investigates the use of emerging non-volatile memory technologies and 3D-FETs to optimize Quality of Service in time-critical systems. His research has been published in several journals such as IEEE TC, IEEE TCAD, IEEE TPDS, ACM TECS, ACM TACO, etc. some conferences including the DAC, IPDPS, CF, DATE, ASAP, etc. 

[FATA] MSci Talks

Group: Formal Analysis, Theory and Algorithms (FATA)
Speaker: Cara Lowe and Duncan Mather
Date: 28 April, 2026
Time: 15:00 - 16:00
Location: Room 422, SAWB

[FATA] MSci Talks

Group: Formal Analysis, Theory and Algorithms (FATA)
Speaker: Andrei Boghean and Josh Duffy
Date: 05 May, 2026
Time: 15:00 - 16:00
Location: Room 422, SAWB

Theory Day 2026

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 07 May, 2026
Time: 01:00 - 01:00
Location: University of Strathclyde, United Kingdom

Join us on 7 May 2026 for Theory Day, bringing together researchers from across Scotland working on Theory and adjacent topics. Staff and students are encouraged to register and submit a proposal to be included in the programme.

Pallas: A Data-Plane-Only Approach to Accurate Persistent Flow Detection on Programmable Switches in High-Speed Networks

Group: Networked Systems Research Laboratory (NETLAB)
Speaker: Weihe Li, University of Edinburgh
Date: 07 May, 2026
Time: 11:00 - 12:00
Location: Sir Alwyn Williams Building, 423 Seminar Room

Abstract:

In high-speed data center networks, persistent flows are repeatedly observed over extended periods, potentially signaling threats such as stealthy DDoS or botnet attacks. Monitoring every flow in production-grade hardware switches that feature limited memory, however, is challenging under typical high flow rates and data volumes. To tackle this, approximate data structures, like sketches, are often employed. Yet many existing methods rely on per-time-window flag resets, which require frequent control-plane interventions that make them unsuitable for high-speed traffic. This paper introduces Pallas, a fully data-plane-implementable sketch for detecting persistent flows in high-speed networks with high accuracy, obviating the need for time-window-based resets. We further propose Opt-Pallas, an enhanced variant of Pallas that improves detection accuracy by incorporating flow arrival patterns. We present a rigorous error bound analysis for both Pallas and Opt-Pallas, along with extensive performance evaluations using a P4-based prototype on an Intel Tofino switch. Pallas scales persistent flow detection to line-rate capacity, while state-of-the-art solutions fail to operate beyond a few Mbps. Our results show that Pallas and Opt-Pallas can accurately detect persistent flows in traffic volumes over 60× higher than those handled by the best existing approach. Additionally, even under low-speed traffic, Pallas and Opt-Pallas achieve 4.21% and 7.85% higher lookup accuracy while consuming only 8.5% and 9.7% of switch resources, respectively. Extensive trace-driven results on a CPU platform further validate the high detection accuracy of Opt-Pallas compared to existing methods.

Bio:

Weihe Li is a postdoctoral researcher at the University of Edinburgh. He received his Ph.D. in Informatics from the University of Edinburgh in 2025. His research focuses on the design of approximate data structures for fast and accurate flow detection in high-speed networks, with an emphasis on practical deployment on programmable switches. His first-author work has appeared in top-tier venues, including ACM SIGMOD 2025, ACM WWW 2024 and 2025 (Best Student Paper Award, 2024), IEEE ICNP 2025, IEEE/ACM Transactions on Networking (ToN), and IEEE Transactions on Computers (TC). He has also conducted research in related networking areas, including video streaming and load balancing in data center networks.

"Abuse Risks are Often Inherent to Product Features": Exploring AI Vendors' Bug Bounty and Responsible Disclosure Policies

Group: Networked Systems Research Laboratory (NETLAB)
Speaker: Yangheran Piao, University of Edinburgh
Date: 21 May, 2026
Time: 11:00 - 12:00
Location: Sir Alwyn Williams Building, 423 Seminar Room

Abstract:

As vendors adopt AI technologies, security researchers are working to uncover and fix related vulnerabilities, which is important given AI systems handle sensitive data and critical functions. This process relies on vendors receiving and rewarding AI vulnerability reports. To assess current practices, we analyzed the vulnerability disclosure policies of 264 AI vendors. We employed a mixed-methods approach, combining snapshot and longitudinal qualitative analysis, as well as comparing alignment with 320 AI incidents and 260 academic articles. Our analysis reveals that 36% of AI vendors have no established policy, and only 18% mention AI risks. Data access, authorization, and model extraction vulnerabilities are most consistently declared in-scope. Jailbreaking and hallucination are most commonly declared out-of-scope. We identify three profiles that reflect vendors' different positions toward AI vulnerabilities: proactive clarification (n = 46), silent (n = 115), and restrictive (n = 103). Our alignment results suggest that vendors may address AI vulnerability disclosure later than academic research and real-world incidents.

Bio:

Yangheran (Lawrence) Piao is a third-year PhD student at the School of Informatics, University of Edinburgh. His research explores usable security, security economics, and cybercrime, with a specific focus on the vulnerability disclosure ecosystem, bug bounties, and AI vulnerability reporting. Yangheran’s work has been published and presented at premier security venues, including USENIX Security, IEEE S&P (Oakland), and WEIS.

Scottish Argumentation Day 2026

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 29 May, 2026
Time: 01:00 - 01:00
Location: University of Dundee

Scotland has a particularly high concentration of research groups working in the AI subfield of computational argumentation. Scottish Argumentation Day has previously been attended by researchers based both in Scotland and further afield, and has enabled the Scottish argumentation community to present their work in an informal setting, share feedback, and strengthen professional links. SAD began with Aberdeen 2011, and most recently took place in Edinburgh 2024. In continuing this series, our aim is threefold: (i) enable Scottish argumentation researchers, and especially PhD students, to mutually present their work; (ii) affirm Scottish argumentation research as a recognisable presence; (iii) provide a concrete opportunity for Scottish researchers to network. At SAD 2026 we aim to improve visibility for Scotland-based researchers, especially PhD students and early-career researchers, to encourage knowledge- and skill-exchange at all levels, and to foster cross-institution relations and collaborations. Participation is free but registration is required. ————————————————– We invite abstracts of up to 250 words to be presented as a short talk or poster. Participants at all levels are encouraged to present work, so that everyone can come away with a view of the current Scottish argumentation landscape. We invite abstracts at a range of levels, including: Overview of a specific research project or a lab’s area of work Recent work Work in progress, recent findings or initial results PhD projects and project plans PhD students are especially encouraged to present their projects and project plans to benefit from wider feedback in a supportive atmosphere. Abstract submission form: https://forms.gle/qCVGqi1sahCKATJv6 ————————————————– The day will be scaffolded by three keynote talks by John Lawrence of the University of Dundee, Elena Musi of the University of Liverpool and Henning Wachsmuth of Leibniz University Hannover. ————————————————–

TBC

Group: Networked Systems Research Laboratory (NETLAB)
Speaker: Jinming Yang
Date: 04 June, 2026
Time: 11:00 - 12:00
Location: Sir Alwyn Williams Building, 423 Seminar Room

EASE 2026: International Conference on Evaluation and Assessment in Software Engineering

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 09 June, 2026
Time: 01:00 - 01:00
Location: James McCune Smith Learning Hub, University Avenue, Glasgow, G12 8QW

EASE is an internationally leading venue for academics and practitioners to present and discuss their research on evidence-based software engineering, and its implications for software practice. EASE is ranked as A conference in CORE. The 30th edition of EASE will take place in Glasgow, Scotland. EASE 2026 welcomes high-quality submissions, describing original and unpublished research for the following tracks: full research papers, short papers & emerging results, industry, posters & vision, journal-first, and a doctoral symposium. There will also be co-located events, including workshops and tutorials, and a track planned for journal-first presentations. See conference website for submission tracks and deadlines. EASE 2026

S3CIX 2026 - Symposium and Summer School on Computational Interaction

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 16 June, 2026
Time: 01:00 - 01:00
Location: Sir Alwyn Williams Building, University of Glasgow, 18 Lilybank Gardens, Glasgow, G12 8QN, United Kingdom

Registration for the 10th Symposium and Summer School on Computational Interaction will open 1 February and close 14 March 2026. View programme, event details and registration process at S³CIX 2026. This year S³CIX is expanding from a Summer School format to also include a 4 day long academic Symposium. We anticipate about 30 students and 40 academics and invited speakers to attend. There will also be two workshops. Computational interaction often involves elements from machine learning, signal processing, information theory, optimisation, inference, control theory and formal modelling. Computational interaction would typically involve at least one of: an explicit mathematical model of user-system behaviour; a way of updating that model with observed data from users; an algorithmic element that, using this model, can directly synthesise or adapt the design; a way of automating and instrumenting the modelling and design process; the ability to simulate or synthesise elements of the expected user-system behaviour.”

TBC

Group: Networked Systems Research Laboratory (NETLAB)
Speaker: Muhammad Arif
Date: 18 June, 2026
Time: 11:00 - 12:00
Location: Lilybank Gardens, F121 Conference Room

10th Summer School and Symposium on Computational Interaction (S³CIX)

Group: Inference, Dynamics and Interaction (IDI)
Speaker: multiple
Date: 20 June, 2026
Time: 09:00 - 16:00
Location: Sir Alwyn Williams Building, 422 Seminar Room

Welcome to the Symposium and Summer School on Computational Interaction! This year we are expanding from a Summer School format to also include a 4 day long academic Symposium. We anticipate about 30 students and 40 academics and invited speakers to attend. There will also be two workshops.

SPLV’26: Scottish Programming Languages and Verification Summer School 2026

Group: Scottish Informatics and Computer Science Alliance (SICSA)
Speaker: SICSA Event, SICSA
Date: 03 August, 2026
Time: 01:00 - 01:00
Location: TBA

The 2026 edition of SPLV will be held at the University of Glasgow, with the main courses running from within the Gilbert Scott Building. The school is aimed at PhD students in programming languages, verification and related areas. Researchers and practitioners are welcome, as are strong undergraduate and masters students with the support of a supervisor. Participants should have a background in computer science, mathematics or a related discipline. Prospective students may contact the organisers if they have any concerns about background knowledge. Registration will open March 2026. View full programme at SPLV 2026 | SPLV