[Question] A new research direction for EvoAgentX

[rawathemant246]

@fangjy6 @X-iZhang

i have question currently EvoAgentX framework is very well thought process you guys have worked on but the core problem with this
--> unconstrained agent evolution(Exploration) leads to behavioral drift, hallucination, and cost blowup

What i'm thinking we can introduce a new formulation over EvoAgentX

Constrained Evolutionary Agent Optimization (CEAO)

• training the mutation policy with DPO/GRPO so the mutator learns to propose better offspring can be a novel in this, and that way we can constrain the cost blownup issue, hallucination rate while agents mutate, and behavioural drift

What's your thought in this can we have a talk ? i want to know i want to work on this with you guys

Additional Context (Optional)

No response