[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-117634-en":3,"doc-seo-117634-105":30,"detail-sidebar-cat-0-en-105":91},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":27,"seo_description":14,"update_tm":28,"read_time":29},117634,962075114101,"Seraphina","https://ap-avatar.wpscdn.com/avatar/e000253a75eb197efd?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780044092746381165",8,"Research & Report","How Machine Learning Models Encode Knowledge - and What We Can Learn from Them - Doctor of Philosophy Thesis","A Doctor of Philosophy thesis examining how machine learning models represent and encode knowledge, with a focus on what interpretability and analysis reveal about model understanding. The work is grounded in research practices developed during doctoral study at Oxford, supported through collaborations with leading academic and industry research teams. It emphasizes mentorship, questioning ideas, and building independent research capacity, culminating in insights drawn from collaborative efforts across multiple institutions and research labs.","How Machine Learning Models Encode Knowledge – and What We Can Learn  \nfrom Them  \nLisa Miou Antoinette Schut  \nLinacre College University of Oxford  \nA thesis submitted for the degree of Doctor of Philosophy  \nTrinity 2025  \nThis thesis is dedicated to my family – Han, Heleen, and Donna. Thank you for  \nyour love and support, always.  \nAcknowledgements  \nThroughout my DPhil and time at Oxford, I’ve met a lot of wonderful people who have influenced my life and academic journey. I believe that all the conversations and interactions over time have shaped me as a researcher, for which I’m very grateful.  \nMentors and Collaborators. I would like to thank my supervisor, Yarin Gal. You took a chance on me years ago and have supported me ever since. Thankyou for teaching me how to do research, develop and question ideas, and grow into an independent researcher.  \nBeen Kim – thank you for your mentorship during my internship and beyond. You taught me so much about interpretability, and the field more generally.  \nThroughout my DPhil, I was lucky to collaborate and learn from many incredible researchers. I learned a lot while working with you, and this has only become more clear with time. Thank you for helping me develop my thoughts and ideas, and for helping me grow as a researcher.  \nIn no particular order, I’d like to thank: Tom McGrath, Demis Hassabis, Tom Zahavy, Vivek Veeriah, Shaobo Hou, Kevin Waugh, Matthew Lai, Edouard Leurent, Nenad Tomašev, Satinder Singh, Clare Lyle, Robin Ru, Sebastian Farquhar, Pascal Notin, Angus Nicolson, Alison J. Noble, Jan M. Brauner, Albert Q. Jiang, Benedikt Höltgen, Miroslav Fil, Jiatong Han, Jannik Kossen, Muhammed Razzak, Shreshth A. Malik, Mark van der Wilk, Edward Hu, Greg Yang, Angus Nicolson, Alison Noble, Evan M. Russek, Ionatan Kuperwajs, Yotam Sagiv, Veniamin Veselovsky, Benedikt Stroebl, Gianluca Bencomo, Dilip Arumugam, Arvind Narayanan, Marcelo G. Mattar, Wei Ji Ma, and Thomas L. Griffiths, Oscar Key, Rory McGrath, Luca Costabello, Bogdan Sacaleanu, Medb Corcoran.  \nInstitutional. I am grateful to be supported by the Engineering and Physical Sciences Research Council, and DeepMind.  \nThank you to my examiners, Prof. Max van Kleek and Prof. Bau, for your time and constructive feedback.  \nResearch Labs, Teams and Departments. I was fortunate to visit several labs and teams during my DPhil.  \nAt DeepMind, thank you to Ulrich Paquet and the DEBL team for hosting me. I met a lot wonderful researchers and friends while there – and in particular, I’d like to thank John Hewitt, Bobby He, Kayo Yin, Yash Chandak, and Jordan Hoffmann.  \nAt Google Brain, thank you to Been Kim and the Deep Understanding team. In particular, thank you to Emily Reif, Robert Geirhos, Zi Wang, Greg Yauney, and Blair Bilodeau. I loved coming to the office and chatting with everyone about research.  \nAt Princeton, thank you to the CoCoSci lab and Tom Griffiths for hosting me. I loved spending time at your lab and learning more about Cognitive Science. I am grateful to the friends I’ve made – Ionatan Kuperwajs, Sally Xie, Ratchit Dubey and Cameron Turner – you are (will be) wonderful professors.  \nAt Oxford, I owe a special thanks to the entire OATML lab. I loved working and learning alongside you all. In particular, I am grateful to Lewis Smith, Clare Lyle, Seb Farquhar, and Pascal Notin for your mentorship and guidance during my DPhil.  \nI would also like to thank my AIMS cohort, especially Freddie Bickford Smith, Charig Yang, and Baskaran Sripathmanathan, for your friendship over the years. Thank you to Wendy Poole for your continuous support and help throughout my time here.  \nFinally, I would like to thank my friends in the Department of Statistics – my first academic home at Oxford – Emilien Dupont, Stacy Pu, Fan Wu, and JF Ton.  \nPersonal. To all my friends – thank you for the happiness and support you’ve brought during these years. I feel so lucky to have you in my life.  \nMax, I’m so grateful for our friendship ","cbCain7aPAohQOtf","https://ap.wps.com/l/cbCain7aPAohQOtf","pdf",40204982,1,311,"English","en",105,"# Acknowledgements\n## Mentors and Collaborators\n## Institutional Support and Examiners\n## Research Labs, Teams and Departments\n## Personal Dedications\n## Epilogue","[{\"question\":\"What is the thesis topic?\",\"answer\":\"The thesis explores how machine learning models encode knowledge and what can be learned from analyzing them.\"},{\"question\":\"Who provided key academic support and mentorship?\",\"answer\":\"The acknowledgements highlight the supervisor Yarin Gal and mentorship from Been Kim, along with guidance from colleagues in Oxford research labs.\"},{\"question\":\"Which institutions and research teams are mentioned as collaborators or hosts?\",\"answer\":\"The text thanks DeepMind and Google Brain teams, as well as the CoCoSci lab at Princeton and the OATML lab and Department of Statistics at Oxford.\"}]","How Machine Learning Models Encode Knowledge - and What We Can Learn from Them - Doctor of Philosophy Thesis | PDF",1785677490,784,{"code":4,"msg":31,"data":32},"ok",{"site_id":24,"language":23,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"how-machine-learning-models-encode-knowledge-and-what-we-can-learn-from-them-doctor-of-philosophy-thesis","",{"@graph":36,"@context":85},[37,54,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/how-machine-learning-models-encode-knowledge-and-what-we-can-learn-from-them-doctor-of-philosophy-thesis/117634/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":23,"description":14,"dateModified":62,"datePublished":62,"encodingFormat":61,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-08-02",true,{"@type":65,"interactionType":66,"userInteractionCount":4},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What is the thesis topic?","Question",{"text":75,"@type":76},"The thesis explores how machine learning models encode knowledge and what can be learned from analyzing them.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"Who provided key academic support and mentorship?",{"text":80,"@type":76},"The acknowledgements highlight the supervisor Yarin Gal and mentorship from Been Kim, along with guidance from colleagues in Oxford research labs.",{"name":82,"@type":73,"acceptedAnswer":83},"Which institutions and research teams are mentioned as collaborators or hosts?",{"text":84,"@type":76},"The text thanks DeepMind and Google Brain teams, as well as the CoCoSci lab at Princeton and the OATML lab and Department of Statistics at Oxford.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]