Orchestrator should when available, be able to choose per agent model. This would enhance workflows with different models in addition to various roles and prompts.
Multiple papers point verifiers effectiveness being reliant on model heterogeneity, different model being the only way to reach different attractors shape for reasoning.
Orchestrator should when available, be able to choose per agent model. This would enhance workflows with different models in addition to various roles and prompts.
Multiple papers point verifiers effectiveness being reliant on model heterogeneity, different model being the only way to reach different attractors shape for reasoning.