BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:0738f0aef82955349ace7ea0b6944542
DTSTAMP:20260905T042530Z
SUMMARY:CS MSc Thesis Presentation 15 September 2026
DESCRIPTION:Contact: birger.swahn@cs.lth.se\n\nTuesday\, 15 September there
  will be a master thesis presentation in Computer Science at Lund Universi
 ty\, Faculty of Engineering.The presentation will take place in E:4130 (Lu
 cas).Note to potential opponents: Register as an opponent to the presentat
 ion of your choice by sending an email to the examiner for that presentati
 on (firstname.lastname@cs.lth.se). Do not forget to specify the presentati
 on you register for! Note that the number of opponents may be limited (oft
 en to two)\, so you might be forced to choose another presentation if you 
 register too late. Registrations are individual\, just as the oppositions 
 are! More instructions for opponents are found here on the LTH thesis proj
 ect page.13:00-14:00 in E:4130 (Lucas)Presenter: Zhehao ChenTitle: Transfo
 rmer-based Normalizing Flow for Robot Arm Control---RoboTarFlowExaminer: D
 avide TateoSupervisors: Volker Krueger (LTH)This thesis investigates wheth
 er a transformer-based autoregressive normalizing flow can serve as a visu
 omotor policy for bimanual robot manipulation. We adapt the TarFlow image-
 generation architecture so that each token represents a timestep of a join
 t command rather than an image patch\, and condition the flow on camera ob
 servations via cross-attention. The resulting policy\, RoboTarFlow\, is ev
 aluated on all 50 RoboTwin 2.0 Easy tasks against two baselines: NF-P and 
 Diffusion Policy. Ablation studies isolate the design choices that matter:
  the prediction horizon and frame stride dominate\, followed by the vision
  encoder\, where a self-supervised transformer backbone DINOv2 clearly out
 performs ResNet18\, and by the training action noise. The results indicate
  that autoregressive normalizing flows are a viable alternative to diffusi
 on-based policies\, reaching competitive performance while retaining exact
  likelihood computation.&nbsp\;\n\nMore information about the event: https
 ://www.cs.lth.se/en/calendar/cs-msc-thesis-presentation-15-september-2026
DTSTART;TZID=GMT:20260915T110000
DTEND;TZID=GMT:20260915T120000
LOCATION:E:4130 (Lucas)
END:VEVENT
END:VCALENDAR