IROS 2026 has 1,933 papers; researcher filters them to 130 must-reads on robot learning
GlenBerseth · x · 2026-09-28
IROS 2026 in Pittsburgh accepted 1,933 papers. Researcher Glen Berseth filtered the program down to his group's focus areas — robot learning, VLA models, RL fine-tuning of robot policies, imitation learning, diffusion/flow policies, and world models — into 3 reading lists totaling 130 papers:
- P0 must-read: 38 papers; P1 should-read: 69; P2 nice-to-read: 23
- Each entry has a one-line summary from the abstract plus session time and room
- Papers with arXiv preprints link directly; others link to IEEE Xplore search
P0 features several VLA generalization works, such as preserving pretrained representations in VLA fine-tuning, EgoVLA learning from egocentric human videos, and cross-embodiment VLA pretraining from human EgoVideos via latent actions.
Related event: IROS 2026 Accepts 1,933 Papers; Researcher Curates 130-Paper Reading List(2 posts)→
More from Embodied
- Tesla workers push back on mocap suits for Optimus training, sparking data-royalty debate — VraserX · 2026-09-29
- Harness device hands-on: a $149 desk command center for all your AI agents — dee_hw · 2026-09-29
- Letting Codex Build Robotics Work: Simulation Was Key, But It Maxes Out Line Limits — zeeg · 2026-09-29
- "Never Seen a Robot Clean a Actually Dirty House": Viral Robot Cleaning Demo Draws Backlash — VoidStateKate · 2026-09-29
- DynaRobotics teases first look at new robot built for real-world work — Rewkang · 2026-09-29
- IROS 2026 has 1,933 papers — researcher curates 130-paper reading list on VLA and robot learning — GlenBerseth · 2026-09-28