What it asks
A research programme at Deakin University’s Applied Artificial Intelligence Institute (A²I²) studies how to route each task to the cheapest model that is sufficient for it, while the models drift and feedback is sparse. Project
RouteLLM, a 2024 paper, describes routers that choose at inference time between a stronger and a weaker LLM, balancing cost against the quality of the response.
How it is studied
The programme studies routing under two conditions, model drift and sparse feedback. Its methods, data and models are not described here yet.
Where it stands
Ongoing, as of October 2026. Project No result is claimed here.
My part
I contribute to the programme.
