The tweet was deleted by the author.
But we saved everything 🙂.
Tanishq Mathew Abraham has raised concerns over the pace of development in the LLM RL ecosystem. Despite 9 months elapsing since the release of R1, he believes the progress in the open-source community should be more advanced.
Abraham agrees with Rohan's conclusion that one of the top libraries currently available is prime-rl. The expectation is for the open-source sector to quickly enhance the robustness and efficiency of these machine learning frameworks.
Abraham's critique of the current trajectory in open-source RL resonates with his earlier efforts to improve data efficiency, as seen when he introduced DEPO for reinforcement learning. His ongoing evaluation of the sector also builds upon prior reflections concerning the sustained relevance of AI coding tools, underscoring a consistent call for innovation and robustness in the broader machine learning landscape.