http://talks.cam.ac.uk/show/rss/5408

The collapse of the Kondo regime, followed by a delayed THz pulse emission upon its recovery has been observed in recent THz spectroscopy experiments on heavy-fermion compounds such as CeCu$$Au${x}$. This work provides a theoretical framework to describe the non-equilibrium dynamics by employing the Anderson lattice model, time-dependent non-equilibrium dynamical mean-field theory, and the non-crossing approximation. We identify two key non-equilibrium mechanisms that play pivotal roles in the collapse and subsequent revival of Kondo coherence. First due to the pulse intensity the hybridization between localized $f$-electrons and conduction electrons increases, shifting the system to a mixed-valence regime, leading to a rapid destruction of the Kondo state. Second, while the distribution function and the single-particle peak recover quickly, the Kondo peak requires significantly more time due to the intrinsic low-energy effects associated with Kondo physics. Additionally, we confirm the system’s ability to emit a non-superradiant delayed pulse upon the recovery of the Kondo regime, confirming the many-body origin of the experimentally observed delayed pulse.

• Speaker: Michael Turaev, University of Bonn
• Monday 09 June 2025, 11:00-12:00
• Venue: Seminar Room 3, RDC.
• Series: Theory of Condensed Matter; organiser: Gaurav.

Add to your calendar or Include in your list

This talk will chart the evolution of vision-language models (VLMs) and illustrate how architectural innovations and training paradigms have progressively closed the gap between visual perception and natural‐language understanding. I will cover models such as CLIP , Flamingo and LLaVA and discuss each of their design principles, strengths and weaknesses, and comparative performance across standard benchmarks.

Teams link available upon request (it is sent out on our mailing list, eng-mlg-rcc [at] lists.cam.ac.uk). Sign up to our mailing list for easier reminders via lists.cam.ac.uk.

• Speaker: Varun Jain (University of Cambridge)
• Wednesday 11 June 2025, 11:00-12:30
• Venue: Cambridge University Engineering Department, CBL Seminar room BE4-38..
• Series: Machine Learning Reading Group @ CUED; organiser: .

Add to your calendar or Include in your list

This talk will chart the evolution of vision-language models (VLMs) and illustrate how architectural innovations and training paradigms have progressively closed the gap between visual perception and natural‐language understanding. I will cover models such as CLIP , Flamingo and LLaVA and discuss each of their design principles, strengths and weaknesses, and comparative performance across standard benchmarks.

Teams link available upon request (it is sent out on our mailing list, eng-mlg-rcc [at] lists.cam.ac.uk). Sign up to our mailing list for easier reminders via lists.cam.ac.uk.

• Speaker: Varun Jain (University of Cambridge)
• Wednesday 11 June 2025, 11:00-12:30
• Venue: Cambridge University Engineering Department, CBL Seminar room BE4-38..
• Series: Machine Learning Reading Group @ CUED; organiser: .

Add to your calendar or Include in your list

The long-period radio transients are a newly-discovered class of Galactic radio sources that produce pulsed emission lasting tens of seconds to several minutes, repeating on timescales of tens of minutes to hours. Such cadence is unprecedented, and there is currently no clear emission mechanism or progenitor that can explain the observations, which include complex polarisation behaviour, pulse microstructure, and activity windows that range from hours to decades.

Could they be ultra-long period magnetars, and connected to the phenomenon of Fast Radio Bursts? Could they be white dwarf pulsars, defying the expectations of the magnetic field evolution of these stellar remnants? In this talk I will describe the ten discoveries made so far, informative simulations of their evolution, the potential physical explanations, and the prospects for detecting more of these sources in ongoing and upcoming radio surveys, that will help uncover their true nature.

• Speaker: Prof. Natasha Hurley-Walker (Curtin University)
• Thursday 12 June 2025, 14:00-15:00
• Venue: Coffee area, Battcock Centre.
• Series: Hills Coffee Talks; organiser: Charles Walker.

Add to your calendar or Include in your list

The long-period radio transients are a newly-discovered class of Galactic radio sources that produce pulsed emission lasting tens of seconds to several minutes, repeating on timescales of tens of minutes to hours. Such cadence is unprecedented, and there is currently no clear emission mechanism or progenitor that can explain the observations, which include complex polarisation behaviour, pulse microstructure, and activity windows that range from hours to decades.

Could they be ultra-long period magnetars, and connected to the phenomenon of Fast Radio Bursts? Could they be white dwarf pulsars, defying the expectations of the magnetic field evolution of these stellar remnants? In this talk I will describe the ten discoveries made so far, informative simulations of their evolution, the potential physical explanations, and the prospects for detecting more of these sources in ongoing and upcoming radio surveys, that will help uncover their true nature.

• Speaker: Prof. Natasha Hurley-Walker (Curtin University)
• Thursday 12 June 2025, 11:15-12:00
• Venue: Martin Ryle Seminar Room, Kavli Institute.
• Series: Hills Coffee Talks; organiser: Charles Walker.

Add to your calendar or Include in your list

Abstract not available

• Speaker: James Jackson
• Friday 13 June 2025, 16:00-17:00
• Venue: Tea Room, Old House.
• Series: Bullard Laboratories Tea Time Talks; organiser: David Al-Attar.

Add to your calendar or Include in your list

Abstract not available

• Speaker: Antonio Pellegrino, Department of Mechanical Engineering, University of Bath
• Thursday 06 November 2025, 15:00-16:00
• Venue: Seminar Room West, Room A0.015, Ray Dolby Centre, Cavendish Laboratory.
• Series: Physics and Chemistry of Solids Group; organiser: Stephen Walley.

Add to your calendar or Include in your list

We introduce MultiBLiMP, a massively multilingual benchmark of linguistic minimal pairs, covering 101 languages, 6 linguistic phenomena and containing more than 120.000 minimal pairs. Our minimal pairs are created using a fully automated pipeline, leveraging the large-scale linguistic resources of Universal Dependencies and UniMorph. MultiBLiMP evaluates linguistic abilities of LLMs at an unprecedented multilingual scale, and highlights the shortcomings of the current state-of-the-art in modelling low-resource languages.

• Speaker: Jaap Jumelet (University of Groningen)
• Friday 13 June 2025, 12:00-13:00
• Venue: ONLINE ONLY. Here is the Zoom link: https://cam-ac-uk.zoom.us/j/4751389294?pwd=Z2ZOSDk0eG1wZldVWG1GVVhrTzFIZz09.
• Series: NLIP Seminar Series; organiser: Suchir Salhan.

Add to your calendar or Include in your list

Neutrino oscillations have the potential to answer some of the highest priority open questions in particle physics, such as why is there a matter-antimatter asymmetry in the universe, and where does the flavour structure of the standard model come from. T2K and NOvA individually have the world’s leading single-experiment precision on several of the parameters of the PMNS neutrino oscillation model. Combining the data from these two experiments not only increases their statistical power, but allows degeneracies present in the individual data sets to be lifted. This talk will describe the results of the first combination of these two data sets.

• Speaker: Patrick Dunne: Imperial College London
• Tuesday 10 June 2025, 11:00-12:00
• Venue: Seminar Room -- RDC D2.002 .
• Series: Cavendish HEP Seminars; organiser: Dr Paul Swallow.

Add to your calendar or Include in your list

In this talk I will present an efficient algorithm for a central problem in quadratic Fourier analysis, and which can be seen as a quadratic generalisation of the celebrated Goildreich-Levin algorithm. More precisely, given a bounded function f on the Boolean hypercube {0, 1}n and any ε > 0, our algorithm returns a quadratic polynomial q: {0, 1}n → {0, 1} so that the correlation of f with the function (−1)q is within an additive ε of the maximum possible correlation with a quadratic phase function. This algorithm runs in O(n3) time and makes O(n2 log n) queries to f. As a corollary, we obtain an algorithmic inverse theorem for the order-3 Gowers norm with polynomial guarantees.

Our algorithm is obtained using ideas from recent work on quantum learning theory. Its construction significantly deviates from previous approaches based on algorithmic proofs of the inverse theorem for order-3 Gowers norms (and in particular does not rely on the recent resolution of the polynomial Freiman-Ruzsa conjecture).

Based on joint work with Jop Briët.

Please note that this talk will exceptionally take place in MR14 .

• Speaker: Davi de Castro Silva (University of Cambridge)
• Wednesday 11 June 2025, 13:30-15:00
• Venue: MR14, CMS.
• Series: Discrete Analysis Seminar; organiser: Julia Wolf.

Add to your calendar or Include in your list

In this talk I hope to show how I applied what I learned at the IEEF in my career as a consulting engineer. Of particular utility to me has been the idea of breaking a complex engineering problem into small tractable pieces. I am obliged to briefly introduce my company, Itasca International, and the type of work we do. I will show three examples: Potash is a water soluble rock made of potassium salts, it is economically important because its use as a fertilizer. In North America, potash is solution mined by circulating water that dissolves the rock. This is a rich problem that involves chemistry, fluid flow, heat transfer, and geomechanics. I will demonstrate some models that are used to help design solution mines, forecast production, and diagnose operational problems. Explosives are an inexpensive means to break and move rock for civil purposes like tunneling, road cut development, and open pit mine excavation. Rock blasting is a complex set of processes that span several orders of magnitude in time-scale, length-scale, and stress magnitude. I will describe some simple mathematical and numerical models that have helped understand blasting. Onshore wind energy is rapidly growing in the United States, partially as a consequence of the Inflation Reduction Act of 2022. During construction, the world’s largest mobile cranes are used to lift the nacelle and blades of turbines. There have been several high profile cases of these large cranes tipping over and being destroyed during construction. It is 2025, so every talk has to have something about machine learning now: I will describe the technical problem of soil bearing capacity failure and show how machine learning, via the concept of a surrogate model, has helped make wind turbine installation faster, safer, and less expensive.

Bio: Jason Furtney was a student at the IEEF from 2002 to 2006 after studying Geology at Edinburgh University. Since leaving the institute, Jason has been working as a consulting engineer for Itasca International, a geomechanics consulting and software company in Minneapolis, Minnesota.

• Speaker: Jason Furtney, Itasca International
• Thursday 03 July 2025, 11:30-12:30
• Venue: Open Plan Area, Institute for Energy and Environmental Flows, Madingley Rise CB3 0EZ.
• Series: Institute for Energy and Environmental Flows (IEEF); organiser: Catherine Pearson.

Add to your calendar or Include in your list

Initial support by the British Empire Cancer Campaign in Uganda led to the description of Burkitt Lymphoma and subsequently to the establishment of the Uganda Cancer Institute (UCI). The Uganda Cancer Institute was established in 1967 as a result of a collaboration between Makerere University, the Ministry of Health and the US National Cancer Institute. It was established as a treatment centre for the then recently discovered Burkitt Lymphoma, and was expanded in 1969 to cater for all cancer. The Institute participated in the seminal initial studies on combination chemotherapy. However, years of political turmoil led to a steady decline in care, research and training. Over the past ten years, the UCI has been building capacity to address the cancer care gap and here we describe some of the steps taken towards this effort. The Institute has expanded clinical care capacity, increased human resource capacity and is currently building a cancer research and innovation facility. The Institute is undertaking high quality research and here we describe how our model could also serve other developing countries in building capacity for cancer care and research to address the growing burden of cancer in LMI Cs.

• Speaker: Dr Nixon Niyonzima, Uganda Cancer Institute
• Thursday 19 June 2025, 13:00-14:00
• Venue: CRUK CI Lecture Theatre.
• Series: Cancer Research UK Cambridge Institute (CRUK CI) Seminars in Cancer; organiser: .

Add to your calendar or Include in your list

Women of Afro-Caribbean descent confront more aggressive breast cancer subtypes at a younger age than their Caucasian counterparts. Yet, breast cancer research and treatment development have predominantly focused on Caucasian populations, neglecting potential biological drivers of these disparities. Our study addresses this gap by in-depth characterising the breast tumour microenvironment (TME) in an ethnically diverse cohort. We analysed treatment-naïve breast cancer samples from 45 Black British and 45 White British women, matched by age, tumour subtype, and stage by employing spatial transcriptomics (NanoString GeoMx) and hyper-plex protein profiling (Leica Microsytems Cell DIVE ). We captured whole-transcriptome data from cancer (PanCK+), immune (CD45+), and stromal (aSMA+) compartments from both tumour centre and tumour edge. The most striking differences emerged within the immune and stromal compartments, not in the cancer cells, underscoring metabolic, adhesion, and extracellular matrix rewiring in Black British tumours. Complementary spatial protein profiling further revealed changes in tissue architecture with distinct recurrent patterns of cellular organisation and cell-cell interactions, involving endothelial and B-cells. Our findings suggest that the TME plays a pivotal role in driving ethnic disparities in breast cancer, highlighting the urgent need for ethnically tailored therapies and more inclusive clinical trials to advance precision cancer care. This breakthrough offers new avenues for improving overall outcomes in breast cancer.

• Speaker: Professor Kairbaan Hodivala-Dilke, Centre for Tumour Microenvironment, 2 Barts Cancer Institute, a Cancer Research UK of Excellence, Queen Mary University of London, Charterhouse Square, London, EC1M 6BQ, UK
• Thursday 10 July 2025, 13:00-14:00
• Venue: CRUK CI Lecture Theatre.
• Series: Cancer Research UK Cambridge Institute (CRUK CI) Seminars in Cancer; organiser: Kate Davenport.

Add to your calendar or Include in your list

How does the brain achieve both stability and flexibility in behavior? In this journal club, we will explore two recent studies that illuminate distinct yet interconnected neural mechanisms underlying long-term motor memory and flexible decision-making. The first paper (Kim et al., Nature, 2024) demonstrates that motor memories are encoded in a combinatorial, context-specific manner in the motor cortex of mice. Using long-term two-photon imaging, the authors show that new motor skills are acquired without overwriting old ones, as new preparatory activity patterns emerge in parallel across contexts—offering a robust mechanism for continual learning.

The second paper (Pagan et al., Nature, 2024) investigates how individual variability shapes context-dependent decision-making in rats. The authors develop a behavioral paradigm and theoretical framework revealing three distinct dynamical strategies for evidence accumulation, all capable of supporting flexible behavior. Strikingly, different individuals express different combinations of these strategies, despite similar performance, highlighting substantial neural and computational diversity.

Optionally, we will also discuss findings from a third study (Mishchanchuk et al., Science, 2024), which reveals how the ventral hippocampus encodes abstract contextual states critical for hidden state inference. This study complements the others by highlighting the importance of hippocampal representations in decision-making based on latent contexts. Together, these studies provide a compelling picture of how the brain balances flexibility and stability through context-specific encoding, diverse computational strategies, and abstract contextual inference—shedding light on the neural basis of learning, memory, and cognition.

• Speaker: Máté Lengyel; Guillaume Hennequin
• Tuesday 10 June 2025, 11:00-12:30
• Venue: CBL Seminar Room, Engineering Department, 4th floor Baker building.
• Series: Computational Neuroscience; organiser: .

Add to your calendar or Include in your list

How does the brain achieve both stability and flexibility in behavior? In this journal club, we will explore two recent studies that illuminate distinct yet interconnected neural mechanisms underlying long-term motor memory and flexible decision-making. The first paper (Kim et al., Nature, 2024) demonstrates that motor memories are encoded in a combinatorial, context-specific manner in the motor cortex of mice. Using long-term two-photon imaging, the authors show that new motor skills are acquired without overwriting old ones, as new preparatory activity patterns emerge in parallel across contexts—offering a robust mechanism for continual learning. The second paper (Pagan et al., Nature, 2024) investigates how individual variability shapes context-dependent decision-making in rats. The authors develop a behavioral paradigm and theoretical framework revealing three distinct dynamical strategies for evidence accumulation, all capable of supporting flexible behavior. Strikingly, different individuals express different combinations of these strategies, despite similar performance, highlighting substantial neural and computational diversity. Optionally, we will also discuss findings from a third study (Mishchanchuk et al., Science, 2024), which reveals how the ventral hippocampus encodes abstract contextual states critical for hidden state inference. This study complements the others by highlighting the importance of hippocampal representations in decision-making based on latent contexts. Together, these studies provide a compelling picture of how the brain balances flexibility and stability through context-specific encoding, diverse computational strategies, and abstract contextual inference—shedding light on the neural basis of learning, memory, and cognition.

• Speaker: Máté Lengyel; Guillaume Hennequin
• Tuesday 10 June 2025, 11:00-12:30
• Venue: CBL Seminar Room, Engineering Department, 4th floor Baker building.
• Series: Computational Neuroscience; organiser: .

Add to your calendar or Include in your list

This Cambridge Immunology Network Seminar will take place on Thursday 12 June 2025, starting at 4:00-5:00pm

Speaker: Professor Thomas Bowden, Wellcome Centre for Human Genetics, Oxford

Title: “Characterizing and preventing host cell entry of emerging RNA viruses”

Host: Yorgo Modis, CITIID , Cambridge

Location: Max Perutz Lecture Theatre, MRC LMB

Refreshments will be available following the seminar.

Please note the change of venue to: Max Perutz Lecture Theatre, MRC LMB

• Speaker: Professor Thomas Bowden, Wellcome Centre for Human Genetics, Oxford
• Thursday 12 June 2025, 16:00-17:00
• Venue: Max Perutz Lecture Theatre, MRC LMB, Francis Crick Avenue, Cambridge CB2 0QH .
• Series: Cambridge Immunology Network Seminar Series; organiser: Ruth Paton.

Add to your calendar or Include in your list

Typically, how much do we know about an uncertain dynamical system (UDS) matters a lot when we want to control them. Aiming to accurately capture the evolution of such UDS is impossible as true system uncertainties cannot be captured exactly. Lack of exact system knowledge increases the difficulty in estimating the limits of the uncertain system’s performance. As a result, we often seek to control such UDS such that the resulting control decisions from Robust Decision Making (RDM) paradigms render the UDS insensitive to what we don’t know about them. However, nature can violate the assumptions that the RDM module assume for the system uncertainties with small probability. Controlling UDS under such unforeseen events necessitate the addition of probabilistic rigour on top of the existing RDM approaches. In this talk, I shall propose a Probabilistic RDM (PRDM) approach using the uncertain gap between the dynamical system models (with and without the uncertainty) induced by appropriate distance metric. The proposed framework will allow us to analyse the potential performance degradation of a control action on an UDS when such rare violation events occur. The fertile nature of the probabilistic robust control research area will be highlighted using a list of interesting future research directions.

The seminar will be held in JDB Seminar Room, Department of Engineering, and online (zoom): https://newnham.zoom.us/j/92544958528?pwd=YS9PcGRnbXBOcStBdStNb3E0SHN1UT09

• Speaker: Venkatraman Renganathan, Cranfield University
• Thursday 12 June 2025, 14:00-15:00
• Venue: JDB Seminar Room, Department of Engineering and online (Zoom).
• Series: CUED Control Group Seminars; organiser: Fulvio Forni.

Add to your calendar or Include in your list

Abstract not available

• Speaker: Allan McRobie, University of Cambridge
• Friday 13 June 2025, 15:00-16:00
• Venue: CivEng Seminar Room (1-33) (Civil Engineering Building).
• Series: Engineering Department Structures Research Seminars; organiser: Shehara Perera.

Add to your calendar or Include in your list

Large Language Models (LLMs) have achieved remarkable fluency and versatility—but understanding how they represent meaning internally remains a challenge. In this talk, we explore the emerging science of latent concepts in LLMs: the semantic abstractions implicitly encoded in their internal activations.

We examine how concepts—such as truthfulness, formality, or sentiment—can be represented as low-dimensional structures, discovered through training dynamics, and understood through the lens of linear algebra and associative memory. We discuss the implications for interpretability, robustness, and control, including how concepts can be steered at test time to adjust model behavior without retraining. Specifically, we explore empirical and theoretical evidence supporting the linear representation hypothesis, where such concepts correspond to vectors or affine subspaces, emerging naturally from training dynamics and next-token prediction objectives. We further show that LLMs behave as associative memory systems, retrieving outputs based on latent similarity rather than logical inference. This behavior underlies phenomena such as context hijacking, where semantically misleading prompts can bias the model’s response.

We introduce formal latent concept models that unify these ideas, describe conditions under which concepts are identifiable, and propose learning algorithms for extracting interpretable, controllable representations. We argue that such latent concept modeling offers a principled framework for bridging representation learning with interpretability and model alignment, and offers a promising path toward safer, more controllable, and more trustworthy AI.

• Speaker: Prof. Pradeep Ravikumar, Carnegie Mellon University
• Tuesday 10 June 2025, 14:00-15:00
• Venue: JDB Seminar Room, CUED.
• Series: Probabilistic Systems, Information, and Inference Group Seminars; organiser: Prof. Ramji Venkataramanan.

Add to your calendar or Include in your list

Abstract not available

• Speaker: Max Ryabinin
• Wednesday 18 June 2025, 15:00-16:00
• Venue: Computer Lab, LT1.
• Series: Cambridge ML Systems Seminar Series; organiser: Sally Matthews.

Add to your calendar or Include in your list