Applying and Verifying a Small Language Model on Inexpensive Hardware

We applied the Phi3 Mini SLM on a Raspberry Pi 5 in two domains - language translation and robot control. To compensate for hallucinations, we included a feedback loop in both applications. Although Phi3 Mini runs slowly, it suffices to test these proof-of-concept applications. In testing our applications, we found that the feedback loop enabled users to compensate to some extent for the limitations of the SLM.