James Oswald, Daniel Oblinsky, Volodymyr Varha, Vasilije Dragovic, Harsha Kokel, Kavitha Srinivas, Michael Katz, Shirin Sohrabi
View original ↗Create a feedback-loop framework where an agent searches for planning domains by iterative trial and error against execution feedback. It bridges the gap between 'good description' and 'functional domain'.
Suggested repo: plan-search
"Generate reliable planning domains through agentic self-reflection."
Estimated effort: 100h