Your Question
I noticed that the GEPA optimizer in the DSPy project is reported to outperform reinforcement learning–based approaches, so I was interested in evaluating its effectiveness.
However, when checking the repository, I found that main is about 25 commits ahead of feature/gepa, while feature/gepa is only 1 commit ahead of main. This made me wonder why the GEPA branch has not been merged into main, given that it represents a newer optimization algorithm.
Could you share the reason behind this? For example, is it due to stability concerns, ongoing experiments, API changes, or other design considerations?
Thanks in advance for the clarification.
Additional Context (Optional)
No response
Your Question
I noticed that the GEPA optimizer in the DSPy project is reported to outperform reinforcement learning–based approaches, so I was interested in evaluating its effectiveness.
However, when checking the repository, I found that main is about 25 commits ahead of feature/gepa, while feature/gepa is only 1 commit ahead of main. This made me wonder why the GEPA branch has not been merged into main, given that it represents a newer optimization algorithm.
Could you share the reason behind this? For example, is it due to stability concerns, ongoing experiments, API changes, or other design considerations?
Thanks in advance for the clarification.
Additional Context (Optional)
No response