The programme · robotics paper reading clubs
Find your reading club. Take a paper. Present it in ten.
Robotics research outgrew any one reader, so labs started reading in rooms: one person takes a paper apart, everyone else gets the signal in ten minutes, and the room argues about what it means. This is the index of those rooms, what they read, who presents, and a six-slide kit for any paper you want to bring.
Refreshed twice a week from the clubs' own pages and the arXiv API. Latest full month: 1,169 submissions to cs.RO in July 2026, up 46% on a year earlier.
Next sessions
Bring one of these on Saturday
Papers a club has already put on the table, each with a slide kit pulled from its own sections: the problem, what came before, the method, the numbers, where it breaks, and the one line worth remembering.
Nobody has presented these yet
Ranked for presentability: recent, venue-accepted or already collecting citations, code available. If you need a paper for this week, start here and be the first.
Who is in the room
The mix is the point
A PhD student three years into tactile sensing, an engineer shipping arms to factories, a founder who read the paper on the train, someone switching in from computer vision. Different stages, different backgrounds, the same paper, and healthy disagreement. You can come and just listen. Presenting is open to anyone, and at some clubs it earns you priority when spots run short.
What is moving
Vision-language-action is taking share
Vision-language-action +7.1pp, World models +2.86pp, Tactile & force sensing +2.2pp. Share of monthly cs.RO submissions, last six months against the twelve before. One arXiv query per topic per month, query strings published.
Your city
7 cities are missing one
Tel Aviv, Singapore, London, Tokyo, Bangalore and more have world-class robotics labs and no public reading club we could find. Starting one costs a room, a date and a first presenter, which is usually you.
Browse by what you work on
World models
Learned simulators that predict future observations, so a policy can plan or train inside imagination instead of on hardware.
Vision-language-action
Policies that consume pixels and language and emit robot actions directly, usually fine-tuned from a VLM backbone.
Dexterous manipulation
Multi-fingered hands, in-hand reorientation, and contact-rich control of everyday objects.
Humanoids & locomotion
Whole-body control, legged locomotion, and loco-manipulation on human-form robots.
Egocentric & human data
Turning first-person human video and wearable capture into robot training signal, bypassing teleoperation.
Tactile & force sensing
Touch as a first-class modality: skins, current sensing, and contact-rich feedback loops.
Simulation, sim-to-real & benchmarks
Simulators, domain randomisation, and the evaluation harnesses that decide whether a policy really works.
Imitation & diffusion policies
Behaviour cloning at scale, and the generative-model action heads that made it work.
Reinforcement learning & control
Online and offline RL, model-predictive control, and the safety filters wrapped around them.