Reading · Sarvik

// 30 papers

Vision-language-action & robot learning

The papers I read building VLA-DSS, an efficient RGB-only robot policy.

Octo: An Open-Source Generalist Robot Policy Octo Model Team — Berkeley / Stanford / CMU / DeepMind · 2024
core
OpenVLA: An Open-Source Vision-Language-Action Model Kim, Pertsch, Karamcheti et al. · 2024
out of scope
π₀: A Vision-Language-Action Flow Model for General Robot Control Physical Intelligence — Black, Brown, Driess et al. · 2024
gate
VLA-R1: Enhancing Reasoning in Vision-Language-Action Models Ye, Zhang, Wang et al. — GigaAI / CASIA / Tsinghua · 2025
out of scope
BitVLA: 1-bit Vision-Language-Action Models for Robotics Manipulation Wang, Xiong, Wang, Chen — CAS · 2025
context
HAMLET: Switch Your VLA into a History-Aware Policy Koo et al. — KAIST / UC Berkeley (ICLR 2026) · 2026
context
RT-1: Robotics Transformer for Real-World Control at Scale Brohan et al. — Google · 2022
context
RT-2: Vision-Language-Action Models Transfer Web Knowledge to Robotic Control Brohan et al. — Google DeepMind · 2023
context
Open X-Embodiment: Robotic Learning Datasets and RT-X Models Open X-Embodiment Collaboration · 2023
context

DAgger: A Reduction of Imitation Learning to No-Regret Online Learning Ross, Gordon, Bagnell · 2011
core
SwiftVLA: auxiliary spatial (x-y) prediction for lightweight VLAs 2025
core
Fourier Controller Networks for Real-Time Decision-Making (FCNet) Tan et al. — ICML 2024 · 2024
core
Fourier Neural Operator for Parametric PDEs (FNO) Li et al. · 2020
core
Invariant Scattering Convolution Networks (Wavelet Scattering) Bruna & Mallat; Mallat · 2012–13
core
DINOv3: self-supervised vision backbone Meta AI · 2025
core
SmolVLA: a small VLA baseline Hugging Face · 2025
context

30 papers, one field so far. Write-ups live in the paper notes.