Reading Group

A reference document used to keep track of the reading group meetings. Pull requests are welcome. Meeting Link

Schedule 2025

DatePresenterPaperWhy should we careAppear inRepoBlogs
22/01/25Jooyoung LeeAre Large Language Models Capable of Generating Human-Level Narratives?Analyses AI generated storylines through three discourse-level aspects: i) story arcs; ii) turning points; and iii) affective dimensions including arousal and valenceEMNLP'24
05/02/25Piotr BrodkaInvited talk from our international visitor
19/02/25Elaine GongDeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement LearningDeepSeek's first-generation reasoning models DeepSeek-R1-Zero and DeepSeek-R1arxiv
05/03/25Frankie YuanA primer on partially observable Markov decision processes (POMDPs)A (somewhat) gentle introduction to Partially Observed Markov Decision Processes (POMDP)Methods in Ecology and Evolution
19/03/25Emily BoothUsing thematic analysis in psychologyA basic introduction to one form of qualitative data analysis used widely by social science researchersQualitative Research in Psychology
09/04/25Frankie YuanBehavioral Homophily in Social Media via Inverse Reinforcement Learning: A Reddit Case StudyFrankie will dry-run his WWW presentation.The Web Conference (WWW’25)
16/04/25Marian-Andrei RizoiuAn overview of the thesis submission process, including Candidature assessments and selecting reviewers.
30/04/25No reading group! Find us at WWW’25
07/05/25Philipp SchneiderReinforcement-Learning Based Covert Social Influence OperationsPresents a reinforcement learning framework that models Covert Social Influence Operations on social media to optimise the actions of malicious agents in manipulating public opinion.WWW'25
14/05/25Philipp SchneiderThe AI Economist: Taxation policy design via two-level deep multiagent reinforcement learningIntroduces a two-level deep reinforcement learning framework where economic agents and a social planner co-adapt to design optimal taxation policies.Science
21/05/25Jooyoung LeeExposing Cross-Platform Coordinated Inauthentic Activity in the Run-Up to the 2024 U.S. ElectionIntroduces a coordination detection model to analyse cross-platform coordinated inauthentic activity across X (formerly Twitter) Facebook and Telegram during the 2024 U.S. Election.WWW'25
04/06/25Rohit RamPractical Guidelines for Ideology Detection Pipelines and Psychosocial ApplicationsRohit will dry-run his ICWSM presentationICWSM'25
11/06/25Cancelled
25/06/25Huixiang FuThe susceptibility paradox in online social influenceThe paper investigates susceptibility to influence within social networks focusing on the differential effects of influence-driven versus spontaneous behaviors on user content adoptionICWSM'25
09/07/25Elaine GongHarnessing the Universal Geometry of EmbeddingsAn unsupervised approach translates any embedding to and from a universal latent representationarXiv
23/07/25Lin TianSafety Alignment Should Be Made More Than Just a Few Tokens DeepOutstanding Paper Award at ICLR2025. This paper critically examines why current Large Language Models (LLMs) remain vulnerable despite extensive safety alignment training.ICLR'25
06/08/25Cancelled due to technical issus
20/08/25Frankie YuanA Closer Look at Machine Unlearning for Large Language ModelsExamines the training and evaluation of machine unlearning techniques on LLMs.ICLR'25
03/09/25Lin Tian
10/09/25Cancelled
17/09/25Emily Booth
24/09/25Marian-Andrei Rizoiu
01/10/25Md Zakir HossainInvited lecture by our research visitor.
08/10/25Jooyoung Lee
15/10/25Huixiang Fu
22/10/25Andrew Savchenko
29/10/25Kevin
05/11/25Elaine Gong
12/11/25Frankie Yuan
19/11/25Lin Tian
26/11/25Emily Booth
03/12/25Jooyoung Lee
10/12/25Huixiang Fu

2024 Archive

2023 Archive

2022 Archive

2021 Archive