Seminars

At Helsing, speed and correctness are key in delivering high quality products. The two are often in antithesis; it is difficult to quickly iterate over designs while keeping your codebase correct and vice versa.

In my last talk (https://talks.cam.ac.uk/talk/index/228790), I show four methodological limitations that prevent machine learning systems from reaching the rigour of syllogistic reasoning.

My overarching research question deals with the epigenetic inheritance of disease risk. Numerous multi-generational studies across species have shown that DNA sequence variation only explains a part of the heritability of disease risk.

In natural language, there are usually many ways to say the same thing: the answer to a question can be said multiple ways, and there are many good translations of the same sentence.

Abstract not available

The distinct elements problem in streaming algorithms refers to estimating the number of unique elements in a data stream while using limited memory.

Abstract: As digital technology evolves, our interactions with digital systems are extending beyond traditional visual and auditory interfaces. This talk explores the emerging field of multisensory interactions, where senses such as touch, smell, and even taste are integrated to enrich digital experiences.

Abstract:

The project aims to implement a verified reduction from Temporal Planning to Timed Automata, targeting a model checker which was verified in Isabelle/HOL. The model checker can output and check certificates for unsatisfied reachability properties.

Understanding and mitigating hallucinations in Large Language Models (LLMs) is crucial for ensuring reliable content generation. While previous research has primarily focused on “when” LLMs hallucinate, our work explains “why” and directly links model behaviour to the pre-training data that forms their prior knowledge.