Psi-Robot/Awesome-VLA-Papers

Psi-Robot/Awesome-VLA-Papers is an informational repository serving as the appendix to a survey paper. With 512 stars, it has captured the attention of the robotics research community, particularly those interested in 'Action Tokenization'—a critical bottleneck in scaling robot learning. However, from a competitive intelligence standpoint, its defensibility is near zero as it contains no proprietary code, datasets, or infrastructure. It is a curated markdown list. The primary moat is the academic 'stamp of approval' and the labor of curation, which is easily replicated or superseded by automated discovery tools and RAG-based research assistants (e.g., Consensus, Perplexity). The displacement horizon is short because the VLA field is moving at an extreme velocity; models like Google's RT-2, Physical Intelligence's π0, and OpenVLA are evolving faster than static lists can be maintained. Frontier labs (DeepMind, OpenAI) pose a 'medium' risk not by building a competing list, but by centralizing the VLA ecosystem around their proprietary or semi-open weights and documentation, making third-party surveys less relevant for practitioners. This project is a useful snapshot for academic literature reviews but lacks the technical 'stickiness' or utility of a library like Hugging Face's Lerobot or the Open X-Embodiment datasets.